-
Notifications
You must be signed in to change notification settings - Fork 2
/
Copy path_utils.html
1723 lines (1393 loc) · 151 KB
/
_utils.html
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
<!DOCTYPE html>
<!--[if IE 8]><html class="no-js lt-ie9" lang="en" > <![endif]-->
<!--[if gt IE 8]><!--> <html class="no-js" lang="en" > <!--<![endif]-->
<head>
<meta name="robots" content="noindex">
<meta charset="utf-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>torch._utils — PyTorch 2.2 documentation</title>
<link rel="canonical" href="https://pytorch.org/docs/stable/_modules/torch/_utils.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 Tag Manager -->
<script>(function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':
new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],
j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src=
'https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);
})(window,document,'script','dataLayer','GTM-T8XT4PS');</script>
<!-- End Google Tag Manager -->
<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>
<div id="resourcesDropdownButton" data-toggle="resources-dropdown" class="resources-dropdown">
<a class="resource-option with-down-arrow">
Edge
</a>
<div class="resources-dropdown-menu">
<a class="nav-dropdown-item" href="https://pytorch.org/edge">
<span class="dropdown-title">About PyTorch Edge</span>
</a>
<a class="nav-dropdown-item" href="https://pytorch.org/executorch">
<span class="dropdown-title">ExecuTorch</span>
</a>
</div>
</div>
</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'>2.2 ▼</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">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="../../cpu.html">torch.cpu</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="../../torch_cuda_memory.html">Understanding CUDA Memory Usage</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../torch_cuda_memory.html#generating-a-snapshot">Generating a Snapshot</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../torch_cuda_memory.html#using-the-visualizer">Using the visualizer</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../torch_cuda_memory.html#snapshot-api-reference">Snapshot API Reference</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="../../export.html">torch.export</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="../../torch.compiler.html">torch.compiler</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="../../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="../../utils.html">torch.utils</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="../../deterministic.html">torch.utils.deterministic</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>
<li class="toctree-l1"><a class="reference internal" href="../../logging.html">torch._logging</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._utils</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">
<!-- Google Tag Manager (noscript) -->
<noscript><iframe src="https://www.googletagmanager.com/ns.html?id=GTM-T8XT4PS"
height="0" width="0" style="display:none;visibility:hidden"></iframe></noscript>
<!-- End Google Tag Manager (noscript) -->
<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._utils</h1><div class="highlight"><pre>
<span></span><span class="kn">import</span> <span class="nn">copyreg</span>
<span class="kn">import</span> <span class="nn">functools</span>
<span class="kn">import</span> <span class="nn">sys</span>
<span class="kn">import</span> <span class="nn">traceback</span>
<span class="kn">import</span> <span class="nn">warnings</span>
<span class="kn">from</span> <span class="nn">collections</span> <span class="kn">import</span> <span class="n">defaultdict</span>
<span class="kn">from</span> <span class="nn">contextlib</span> <span class="kn">import</span> <span class="n">nullcontext</span>
<span class="kn">from</span> <span class="nn">typing</span> <span class="kn">import</span> <span class="n">Any</span><span class="p">,</span> <span class="n">DefaultDict</span><span class="p">,</span> <span class="n">List</span><span class="p">,</span> <span class="n">Optional</span>
<span class="kn">import</span> <span class="nn">torch</span>
<span class="k">def</span> <span class="nf">_type</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">dtype</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span> <span class="n">non_blocking</span><span class="o">=</span><span class="kc">False</span><span class="p">,</span> <span class="o">**</span><span class="n">kwargs</span><span class="p">):</span>
<span class="w"> </span><span class="sd">"""Returns the type if `dtype` is not provided, else casts this object to</span>
<span class="sd"> the specified type.</span>
<span class="sd"> If this is already of the correct type, no copy is performed and the</span>
<span class="sd"> original object is returned.</span>
<span class="sd"> Args:</span>
<span class="sd"> dtype (type or string): The desired type</span>
<span class="sd"> non_blocking (bool): If ``True``, and the source is in pinned memory</span>
<span class="sd"> and destination is on the GPU or vice versa, the copy is performed</span>
<span class="sd"> asynchronously with respect to the host. Otherwise, the argument</span>
<span class="sd"> has no effect.</span>
<span class="sd"> **kwargs: For compatibility, may contain the key ``async`` in place of</span>
<span class="sd"> the ``non_blocking`` argument. The ``async`` arg is deprecated.</span>
<span class="sd"> """</span>
<span class="n">non_blocking</span> <span class="o">=</span> <span class="n">_get_async_or_non_blocking</span><span class="p">(</span><span class="s2">"type"</span><span class="p">,</span> <span class="n">non_blocking</span><span class="p">,</span> <span class="n">kwargs</span><span class="p">)</span>
<span class="k">if</span> <span class="n">dtype</span> <span class="ow">is</span> <span class="kc">None</span><span class="p">:</span>
<span class="k">return</span> <span class="bp">self</span><span class="o">.</span><span class="vm">__module__</span> <span class="o">+</span> <span class="s2">"."</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="k">if</span> <span class="nb">isinstance</span><span class="p">(</span><span class="n">dtype</span><span class="p">,</span> <span class="nb">str</span><span class="p">):</span>
<span class="n">dtype</span> <span class="o">=</span> <span class="n">_import_dotted_name</span><span class="p">(</span><span class="n">dtype</span><span class="p">)</span>
<span class="k">if</span> <span class="n">dtype</span> <span class="o">==</span> <span class="nb">type</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span>
<span class="k">return</span> <span class="bp">self</span>
<span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">is_sparse</span><span class="p">:</span>
<span class="k">if</span> <span class="ow">not</span> <span class="n">dtype</span><span class="o">.</span><span class="n">is_sparse</span><span class="p">:</span>
<span class="k">raise</span> <span class="ne">RuntimeError</span><span class="p">(</span><span class="s2">"Cannot cast sparse tensor to dense tensor"</span><span class="p">)</span>
<span class="n">new_module_name</span> <span class="o">=</span> <span class="n">dtype</span><span class="o">.</span><span class="vm">__module__</span><span class="o">.</span><span class="n">replace</span><span class="p">(</span><span class="s2">".sparse"</span><span class="p">,</span> <span class="s2">""</span><span class="p">)</span>
<span class="n">new_values_type_name</span> <span class="o">=</span> <span class="n">new_module_name</span> <span class="o">+</span> <span class="s2">"."</span> <span class="o">+</span> <span class="n">dtype</span><span class="o">.</span><span class="vm">__name__</span>
<span class="n">new_values</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="o">.</span><span class="n">_values</span><span class="p">(</span><span class="bp">self</span><span class="p">)</span><span class="o">.</span><span class="n">type</span><span class="p">(</span><span class="n">new_values_type_name</span><span class="p">,</span> <span class="n">non_blocking</span><span class="p">)</span>
<span class="n">new_indices_type_name</span> <span class="o">=</span> <span class="n">new_module_name</span> <span class="o">+</span> <span class="s2">".LongTensor"</span>
<span class="n">new_indices</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="o">.</span><span class="n">_indices</span><span class="p">(</span><span class="bp">self</span><span class="p">)</span><span class="o">.</span><span class="n">type</span><span class="p">(</span>
<span class="n">new_indices_type_name</span><span class="p">,</span> <span class="n">non_blocking</span>
<span class="p">)</span>
<span class="k">return</span> <span class="n">dtype</span><span class="p">(</span><span class="n">new_indices</span><span class="p">,</span> <span class="n">new_values</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">size</span><span class="p">())</span>
<span class="k">if</span> <span class="n">dtype</span><span class="o">.</span><span class="n">is_sparse</span><span class="p">:</span>
<span class="k">raise</span> <span class="ne">RuntimeError</span><span class="p">(</span><span class="s2">"Cannot cast dense tensor to sparse tensor"</span><span class="p">)</span>
<span class="k">return</span> <span class="n">dtype</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">size</span><span class="p">())</span><span class="o">.</span><span class="n">copy_</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">non_blocking</span><span class="p">)</span>
<span class="k">def</span> <span class="nf">_hpu</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">device</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span> <span class="n">non_blocking</span><span class="o">=</span><span class="kc">False</span><span class="p">,</span> <span class="o">**</span><span class="n">kwargs</span><span class="p">):</span>
<span class="w"> </span><span class="sd">"""Returns a copy of this object in HPU memory.</span>
<span class="sd"> If this object is already in HPU memory and on the correct device, then</span>
<span class="sd"> no copy is performed and the original object is returned.</span>
<span class="sd"> Args:</span>
<span class="sd"> device (int): The destination HPU id. Defaults to the current device.</span>
<span class="sd"> non_blocking (bool): If ``True`` and the source is in pinned memory,</span>
<span class="sd"> the copy will be asynchronous with respect to the host. Otherwise,</span>
<span class="sd"> the argument has no effect.</span>
<span class="sd"> **kwargs: For compatibility, may contain the key ``async`` in place of</span>
<span class="sd"> the ``non_blocking`` argument.</span>
<span class="sd"> """</span>
<span class="n">non_blocking</span> <span class="o">=</span> <span class="n">_get_async_or_non_blocking</span><span class="p">(</span><span class="s2">"hpu"</span><span class="p">,</span> <span class="n">non_blocking</span><span class="p">,</span> <span class="n">kwargs</span><span class="p">)</span>
<span class="n">hpu</span> <span class="o">=</span> <span class="nb">getattr</span><span class="p">(</span><span class="n">torch</span><span class="p">,</span> <span class="s2">"hpu"</span><span class="p">,</span> <span class="kc">None</span><span class="p">)</span>
<span class="k">assert</span> <span class="n">hpu</span> <span class="ow">is</span> <span class="ow">not</span> <span class="kc">None</span><span class="p">,</span> <span class="s2">"HPU device module is not loaded"</span>
<span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">is_hpu</span><span class="p">:</span>
<span class="k">if</span> <span class="n">device</span> <span class="ow">is</span> <span class="kc">None</span><span class="p">:</span>
<span class="n">device</span> <span class="o">=</span> <span class="n">hpu</span><span class="o">.</span><span class="n">current_device</span><span class="p">()</span>
<span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">get_device</span><span class="p">()</span> <span class="o">==</span> <span class="n">device</span><span class="p">:</span>
<span class="k">return</span> <span class="bp">self</span>
<span class="k">else</span><span class="p">:</span>
<span class="k">if</span> <span class="n">device</span> <span class="ow">is</span> <span class="kc">None</span><span class="p">:</span>
<span class="n">device</span> <span class="o">=</span> <span class="o">-</span><span class="mi">1</span>
<span class="k">with</span> <span class="n">hpu</span><span class="o">.</span><span class="n">device</span><span class="p">(</span><span class="n">device</span><span class="p">):</span>
<span class="k">assert</span> <span class="ow">not</span> <span class="bp">self</span><span class="o">.</span><span class="n">is_sparse</span><span class="p">,</span> <span class="s2">"sparse storage is not supported for HPU tensors"</span>
<span class="n">untyped_storage</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">UntypedStorage</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">size</span><span class="p">(),</span> <span class="n">device</span><span class="o">=</span><span class="n">torch</span><span class="o">.</span><span class="n">device</span><span class="p">(</span><span class="s2">"hpu"</span><span class="p">))</span>
<span class="n">untyped_storage</span><span class="o">.</span><span class="n">copy_</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">non_blocking</span><span class="p">)</span>
<span class="k">return</span> <span class="n">untyped_storage</span>
<span class="k">def</span> <span class="nf">_cuda</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">device</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span> <span class="n">non_blocking</span><span class="o">=</span><span class="kc">False</span><span class="p">,</span> <span class="o">**</span><span class="n">kwargs</span><span class="p">):</span>
<span class="w"> </span><span class="sd">"""Returns a copy of this object in CUDA memory.</span>
<span class="sd"> If this object is already in CUDA memory and on the correct device, then</span>
<span class="sd"> no copy is performed and the original object is returned.</span>
<span class="sd"> Args:</span>
<span class="sd"> device (int): The destination GPU id. Defaults to the current device.</span>
<span class="sd"> non_blocking (bool): If ``True`` and the source is in pinned memory,</span>
<span class="sd"> the copy will be asynchronous with respect to the host. Otherwise,</span>
<span class="sd"> the argument has no effect.</span>
<span class="sd"> **kwargs: For compatibility, may contain the key ``async`` in place of</span>
<span class="sd"> the ``non_blocking`` argument.</span>
<span class="sd"> """</span>
<span class="n">non_blocking</span> <span class="o">=</span> <span class="n">_get_async_or_non_blocking</span><span class="p">(</span><span class="s2">"cuda"</span><span class="p">,</span> <span class="n">non_blocking</span><span class="p">,</span> <span class="n">kwargs</span><span class="p">)</span>
<span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">is_cuda</span><span class="p">:</span>
<span class="k">if</span> <span class="n">device</span> <span class="ow">is</span> <span class="kc">None</span><span class="p">:</span>
<span class="n">device</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">cuda</span><span class="o">.</span><span class="n">current_device</span><span class="p">()</span>
<span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">get_device</span><span class="p">()</span> <span class="o">==</span> <span class="n">device</span><span class="p">:</span>
<span class="k">return</span> <span class="bp">self</span>
<span class="k">else</span><span class="p">:</span>
<span class="k">if</span> <span class="n">device</span> <span class="ow">is</span> <span class="kc">None</span><span class="p">:</span>
<span class="n">device</span> <span class="o">=</span> <span class="o">-</span><span class="mi">1</span>
<span class="k">with</span> <span class="n">torch</span><span class="o">.</span><span class="n">cuda</span><span class="o">.</span><span class="n">device</span><span class="p">(</span><span class="n">device</span><span class="p">):</span>
<span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">is_sparse</span><span class="p">:</span>
<span class="n">new_type</span> <span class="o">=</span> <span class="nb">getattr</span><span class="p">(</span><span class="n">torch</span><span class="o">.</span><span class="n">cuda</span><span class="o">.</span><span class="n">sparse</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">indices</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="o">.</span><span class="n">_indices</span><span class="p">(</span><span class="bp">self</span><span class="p">)</span><span class="o">.</span><span class="n">cuda</span><span class="p">(</span><span class="n">device</span><span class="p">,</span> <span class="n">non_blocking</span><span class="p">)</span>
<span class="n">values</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="o">.</span><span class="n">_values</span><span class="p">(</span><span class="bp">self</span><span class="p">)</span><span class="o">.</span><span class="n">cuda</span><span class="p">(</span><span class="n">device</span><span class="p">,</span> <span class="n">non_blocking</span><span class="p">)</span>
<span class="k">return</span> <span class="n">new_type</span><span class="p">(</span><span class="n">indices</span><span class="p">,</span> <span class="n">values</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">size</span><span class="p">())</span>
<span class="k">else</span><span class="p">:</span>
<span class="n">untyped_storage</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">UntypedStorage</span><span class="p">(</span>
<span class="bp">self</span><span class="o">.</span><span class="n">size</span><span class="p">(),</span> <span class="n">device</span><span class="o">=</span><span class="n">torch</span><span class="o">.</span><span class="n">device</span><span class="p">(</span><span class="s2">"cuda"</span><span class="p">)</span>
<span class="p">)</span>
<span class="n">untyped_storage</span><span class="o">.</span><span class="n">copy_</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">non_blocking</span><span class="p">)</span>
<span class="k">return</span> <span class="n">untyped_storage</span>
<span class="k">def</span> <span class="nf">_get_async_or_non_blocking</span><span class="p">(</span><span class="n">function_name</span><span class="p">,</span> <span class="n">non_blocking</span><span class="p">,</span> <span class="n">kwargs</span><span class="p">):</span>
<span class="w"> </span><span class="sd">"""Return the non-blocking flag given the function name and kwargs.</span>
<span class="sd"> Args:</span>
<span class="sd"> function_name (str): the name of the function being used.</span>
<span class="sd"> non_blocking (bool): the default value.</span>
<span class="sd"> **kwargs (dict): the kwargs passed to the function.</span>
<span class="sd"> """</span>
<span class="k">if</span> <span class="ow">not</span> <span class="n">kwargs</span><span class="p">:</span>
<span class="k">return</span> <span class="n">non_blocking</span>
<span class="k">if</span> <span class="nb">len</span><span class="p">(</span><span class="n">kwargs</span><span class="p">)</span> <span class="o">!=</span> <span class="mi">1</span> <span class="ow">or</span> <span class="s2">"async"</span> <span class="ow">not</span> <span class="ow">in</span> <span class="n">kwargs</span><span class="p">:</span>
<span class="n">message</span> <span class="o">=</span> <span class="s2">"</span><span class="si">{}</span><span class="s2">() got an unexpected keyword argument '</span><span class="si">{}</span><span class="s2">'"</span>
<span class="n">argument</span> <span class="o">=</span> <span class="nb">list</span><span class="p">(</span><span class="n">kwargs</span><span class="o">.</span><span class="n">keys</span><span class="p">())</span><span class="o">.</span><span class="n">pop</span><span class="p">()</span>
<span class="k">raise</span> <span class="ne">TypeError</span><span class="p">(</span><span class="n">message</span><span class="o">.</span><span class="n">format</span><span class="p">(</span><span class="n">function_name</span><span class="p">,</span> <span class="n">argument</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">"'async' is deprecated; use 'non_blocking'"</span><span class="p">)</span>
<span class="k">return</span> <span class="n">kwargs</span><span class="p">[</span><span class="s2">"async"</span><span class="p">]</span>
<span class="c1"># Note [Don't serialize hooks]</span>
<span class="c1"># ~~~~~~~~~~~~~~~~~~~~~~~~~~~~</span>
<span class="c1"># Since time immemorial, we have serialized the backward hooks associated with</span>
<span class="c1"># variables. This kind of half-worked--Python can pickle global functions</span>
<span class="c1"># (but not closures!)--but there were problems.</span>
<span class="c1">#</span>
<span class="c1"># - It's fragile. If you serialize a backward hook into a saved</span>
<span class="c1"># model, and then you rename the function associated with the hook,</span>
<span class="c1"># now your saved model is broken and you can't load it anymore.</span>
<span class="c1">#</span>
<span class="c1"># - It's not actually used. The standard recommendation is to</span>
<span class="c1"># serialize the *state_dict* of a model, not the model itself</span>
<span class="c1"># (since this is more stable to code changes affecting the model</span>
<span class="c1"># serialization), and the state dict saves "data" only, thus</span>
<span class="c1"># stripping the backward hooks. In some cases, hooks are</span>
<span class="c1"># essential to the well-functioning of a model (e.g., DDP),</span>
<span class="c1"># but DDP already manages readding the hooks!</span>
<span class="c1">#</span>
<span class="c1"># - We didn't serialize them in many cases. Prior to #10220, we</span>
<span class="c1"># were dropping backward hooks in ForkingPickler. We "fixed" this</span>
<span class="c1"># to be convenient with other serialization sites, but lack of</span>
<span class="c1"># serializing backward hooks wasn't actually the root cause of</span>
<span class="c1"># the bug.</span>
<span class="c1">#</span>
<span class="c1"># With these cases in mind, we have decided that a better strategy</span>
<span class="c1"># is to just NOT serialize hooks at all.</span>
<span class="c1">#</span>
<span class="c1"># Since this is a BC-breaking change, we should warn when we previously</span>
<span class="c1"># serialized a hook, but no longer do so. This will be done by adding a special</span>
<span class="c1"># sentinel property to hooks will be used to suppress this warning. If a hook</span>
<span class="c1"># has the property _torch_serialize_ignore, we will not emit a warning if we</span>
<span class="c1"># attempt to serialize a Tensor with this hook attached to it.</span>
<span class="c1">#</span>
<span class="c1"># By the way, when _backward_hooks is skipped, we must give an EMPTY</span>
<span class="c1"># OrderedDict(), if you pass a None you'll run afoul #12219.</span>
<span class="c1"># TODO: Once we decide to break serialization FC, `storage` no longer needs to</span>
<span class="c1"># be a TypedStorage</span>
<span class="k">def</span> <span class="nf">_rebuild_tensor</span><span class="p">(</span><span class="n">storage</span><span class="p">,</span> <span class="n">storage_offset</span><span class="p">,</span> <span class="n">size</span><span class="p">,</span> <span class="n">stride</span><span class="p">):</span>
<span class="c1"># first construct a tensor with the correct dtype/device</span>
<span class="n">t</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">tensor</span><span class="p">([],</span> <span class="n">dtype</span><span class="o">=</span><span class="n">storage</span><span class="o">.</span><span class="n">dtype</span><span class="p">,</span> <span class="n">device</span><span class="o">=</span><span class="n">storage</span><span class="o">.</span><span class="n">_untyped_storage</span><span class="o">.</span><span class="n">device</span><span class="p">)</span>
<span class="k">return</span> <span class="n">t</span><span class="o">.</span><span class="n">set_</span><span class="p">(</span><span class="n">storage</span><span class="o">.</span><span class="n">_untyped_storage</span><span class="p">,</span> <span class="n">storage_offset</span><span class="p">,</span> <span class="n">size</span><span class="p">,</span> <span class="n">stride</span><span class="p">)</span>
<span class="k">def</span> <span class="nf">get_tensor_metadata</span><span class="p">(</span><span class="n">tensor</span><span class="p">):</span>
<span class="c1"># Tensor's Metadata for serializing.</span>
<span class="c1"># Currently, this only returns a dict[string, bool] specifing whether</span>
<span class="c1"># `conj` or `neg` bit is set.</span>
<span class="k">assert</span> <span class="nb">isinstance</span><span class="p">(</span><span class="n">tensor</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">torch</span><span class="o">.</span><span class="n">_C</span><span class="o">.</span><span class="n">_get_tensor_metadata</span><span class="p">(</span><span class="n">tensor</span><span class="p">)</span> <span class="c1"># type: ignore[attr-defined]</span>
<span class="k">def</span> <span class="nf">set_tensor_metadata</span><span class="p">(</span><span class="n">tensor</span><span class="p">,</span> <span class="n">metadata</span><span class="p">):</span>
<span class="c1"># See `get_tensor_metadata` above</span>
<span class="k">assert</span> <span class="nb">isinstance</span><span class="p">(</span><span class="n">metadata</span><span class="p">,</span> <span class="nb">dict</span><span class="p">)</span>
<span class="k">assert</span> <span class="nb">isinstance</span><span class="p">(</span><span class="n">tensor</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">_C</span><span class="o">.</span><span class="n">_set_tensor_metadata</span><span class="p">(</span><span class="n">tensor</span><span class="p">,</span> <span class="n">metadata</span><span class="p">)</span> <span class="c1"># type: ignore[attr-defined]</span>
<span class="k">def</span> <span class="nf">_rebuild_tensor_v2</span><span class="p">(</span>
<span class="n">storage</span><span class="p">,</span> <span class="n">storage_offset</span><span class="p">,</span> <span class="n">size</span><span class="p">,</span> <span class="n">stride</span><span class="p">,</span> <span class="n">requires_grad</span><span class="p">,</span> <span class="n">backward_hooks</span><span class="p">,</span> <span class="n">metadata</span><span class="o">=</span><span class="kc">None</span>
<span class="p">):</span>
<span class="n">tensor</span> <span class="o">=</span> <span class="n">_rebuild_tensor</span><span class="p">(</span><span class="n">storage</span><span class="p">,</span> <span class="n">storage_offset</span><span class="p">,</span> <span class="n">size</span><span class="p">,</span> <span class="n">stride</span><span class="p">)</span>
<span class="n">tensor</span><span class="o">.</span><span class="n">requires_grad</span> <span class="o">=</span> <span class="n">requires_grad</span>
<span class="k">if</span> <span class="n">metadata</span><span class="p">:</span>
<span class="n">set_tensor_metadata</span><span class="p">(</span><span class="n">tensor</span><span class="p">,</span> <span class="n">metadata</span><span class="p">)</span>
<span class="c1"># NB: This line exists only for backwards compatibility; the</span>
<span class="c1"># general expectation is that backward_hooks is an empty</span>
<span class="c1"># OrderedDict. See Note [Don't serialize hooks]</span>
<span class="n">tensor</span><span class="o">.</span><span class="n">_backward_hooks</span> <span class="o">=</span> <span class="n">backward_hooks</span>
<span class="k">return</span> <span class="n">tensor</span>
<span class="k">def</span> <span class="nf">_rebuild_tensor_v3</span><span class="p">(</span>
<span class="n">storage</span><span class="p">,</span>
<span class="n">storage_offset</span><span class="p">,</span>
<span class="n">size</span><span class="p">,</span>
<span class="n">stride</span><span class="p">,</span>
<span class="n">requires_grad</span><span class="p">,</span>
<span class="n">backward_hooks</span><span class="p">,</span>
<span class="n">dtype</span><span class="p">,</span>
<span class="n">metadata</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span>
<span class="p">):</span>
<span class="n">t</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">tensor</span><span class="p">(</span>
<span class="p">[],</span>
<span class="n">dtype</span><span class="o">=</span><span class="n">dtype</span><span class="p">,</span>
<span class="n">device</span><span class="o">=</span><span class="n">storage</span><span class="o">.</span><span class="n">_untyped_storage</span><span class="o">.</span><span class="n">device</span><span class="p">,</span>
<span class="n">requires_grad</span><span class="o">=</span><span class="n">requires_grad</span><span class="p">,</span>
<span class="p">)</span>
<span class="n">t</span><span class="o">.</span><span class="n">set_</span><span class="p">(</span><span class="n">storage</span><span class="o">.</span><span class="n">_untyped_storage</span><span class="p">,</span> <span class="n">storage_offset</span><span class="p">,</span> <span class="n">size</span><span class="p">,</span> <span class="n">stride</span><span class="p">)</span>
<span class="k">if</span> <span class="n">metadata</span><span class="p">:</span>
<span class="n">set_tensor_metadata</span><span class="p">(</span><span class="n">t</span><span class="p">,</span> <span class="n">metadata</span><span class="p">)</span>
<span class="n">t</span><span class="o">.</span><span class="n">_backward_hooks</span> <span class="o">=</span> <span class="n">backward_hooks</span>
<span class="k">return</span> <span class="n">t</span>
<span class="n">_sparse_tensors_to_validate</span><span class="p">:</span> <span class="n">List</span><span class="p">[</span><span class="s2">"torch.Tensor"</span><span class="p">]</span> <span class="o">=</span> <span class="p">[]</span>
<span class="c1"># In _legacy_load() in serialization.py we unpickle storages after the sparse</span>
<span class="c1"># tensors have been already unpickled. Those storages contain data necessary for</span>
<span class="c1"># validating sparse tensors: indices and values. That's why sparse tensors are</span>
<span class="c1"># first unpickled without any validation, and then this function is called just</span>
<span class="c1"># before _legacy_load() returns, so that all the sparse tensors can be validated</span>
<span class="c1"># in bulk.</span>
<span class="c1">#</span>
<span class="c1"># The same procedure must be followed by _load() in serialization.py because due</span>
<span class="c1"># to Pickler semantics, we have to use the same (non-validating) function for</span>
<span class="c1"># unpickling sparse tensors, regardless of the caller.</span>
<span class="k">def</span> <span class="nf">_validate_loaded_sparse_tensors</span><span class="p">():</span>
<span class="k">try</span><span class="p">:</span>
<span class="k">for</span> <span class="n">t</span> <span class="ow">in</span> <span class="n">_sparse_tensors_to_validate</span><span class="p">:</span>
<span class="k">if</span> <span class="n">t</span><span class="o">.</span><span class="n">layout</span> <span class="ow">is</span> <span class="n">torch</span><span class="o">.</span><span class="n">sparse_coo</span><span class="p">:</span>
<span class="n">torch</span><span class="o">.</span><span class="n">_validate_sparse_coo_tensor_args</span><span class="p">(</span>
<span class="n">t</span><span class="o">.</span><span class="n">_indices</span><span class="p">(),</span> <span class="n">t</span><span class="o">.</span><span class="n">_values</span><span class="p">(),</span> <span class="n">t</span><span class="o">.</span><span class="n">size</span><span class="p">(),</span> <span class="n">t</span><span class="o">.</span><span class="n">is_coalesced</span><span class="p">()</span>
<span class="p">)</span>
<span class="k">elif</span> <span class="n">t</span><span class="o">.</span><span class="n">layout</span> <span class="ow">in</span> <span class="p">{</span>
<span class="n">torch</span><span class="o">.</span><span class="n">sparse_csr</span><span class="p">,</span>
<span class="n">torch</span><span class="o">.</span><span class="n">sparse_csc</span><span class="p">,</span>
<span class="n">torch</span><span class="o">.</span><span class="n">sparse_bsr</span><span class="p">,</span>
<span class="n">torch</span><span class="o">.</span><span class="n">sparse_bsc</span><span class="p">,</span>
<span class="p">}:</span>
<span class="c1"># TODO: Validation currently involves an expensive traversal</span>
<span class="c1"># on CPU, which may include a device transfer.</span>
<span class="k">if</span> <span class="n">t</span><span class="o">.</span><span class="n">layout</span> <span class="ow">in</span> <span class="p">{</span><span class="n">torch</span><span class="o">.</span><span class="n">sparse_csr</span><span class="p">,</span> <span class="n">torch</span><span class="o">.</span><span class="n">sparse_bsr</span><span class="p">}:</span>
<span class="n">compressed_indices</span><span class="p">,</span> <span class="n">plain_indices</span> <span class="o">=</span> <span class="p">(</span>
<span class="n">t</span><span class="o">.</span><span class="n">crow_indices</span><span class="p">(),</span>
<span class="n">t</span><span class="o">.</span><span class="n">col_indices</span><span class="p">(),</span>
<span class="p">)</span>
<span class="k">else</span><span class="p">:</span>
<span class="n">compressed_indices</span><span class="p">,</span> <span class="n">plain_indices</span> <span class="o">=</span> <span class="p">(</span>
<span class="n">t</span><span class="o">.</span><span class="n">ccol_indices</span><span class="p">(),</span>
<span class="n">t</span><span class="o">.</span><span class="n">row_indices</span><span class="p">(),</span>
<span class="p">)</span>
<span class="n">torch</span><span class="o">.</span><span class="n">_validate_sparse_compressed_tensor_args</span><span class="p">(</span>
<span class="n">compressed_indices</span><span class="p">,</span> <span class="n">plain_indices</span><span class="p">,</span> <span class="n">t</span><span class="o">.</span><span class="n">values</span><span class="p">(),</span> <span class="n">t</span><span class="o">.</span><span class="n">size</span><span class="p">(),</span> <span class="n">t</span><span class="o">.</span><span class="n">layout</span>
<span class="p">)</span>
<span class="k">else</span><span class="p">:</span>
<span class="k">raise</span> <span class="ne">NotImplementedError</span><span class="p">(</span>
<span class="sa">f</span><span class="s2">"_validate_loaded_sparse_tensors for layout `</span><span class="si">{</span><span class="n">t</span><span class="o">.</span><span class="n">layout</span><span class="si">}</span><span class="s2">`"</span>
<span class="p">)</span>
<span class="k">finally</span><span class="p">:</span>
<span class="n">_sparse_tensors_to_validate</span><span class="o">.</span><span class="n">clear</span><span class="p">()</span>
<span class="k">def</span> <span class="nf">_rebuild_sparse_tensor</span><span class="p">(</span><span class="n">layout</span><span class="p">,</span> <span class="n">data</span><span class="p">):</span>
<span class="w"> </span><span class="sd">"""</span>
<span class="sd"> Rebuilds a sparse tensor from its sparse storage representation.</span>
<span class="sd"> Args:</span>
<span class="sd"> layout (str): The sparse storage layout of the tensor.</span>
<span class="sd"> data (tuple): The tensor's sparse storage representation.</span>
<span class="sd"> """</span>
<span class="k">if</span> <span class="n">layout</span> <span class="o">==</span> <span class="n">torch</span><span class="o">.</span><span class="n">sparse_coo</span><span class="p">:</span>
<span class="k">if</span> <span class="nb">len</span><span class="p">(</span><span class="n">data</span><span class="p">)</span> <span class="o">==</span> <span class="mi">3</span><span class="p">:</span>
<span class="c1"># For BC:</span>
<span class="n">indices</span><span class="p">,</span> <span class="n">values</span><span class="p">,</span> <span class="n">size</span> <span class="o">=</span> <span class="n">data</span>
<span class="n">is_coalesced</span> <span class="o">=</span> <span class="kc">None</span>
<span class="k">else</span><span class="p">:</span>
<span class="n">indices</span><span class="p">,</span> <span class="n">values</span><span class="p">,</span> <span class="n">size</span><span class="p">,</span> <span class="n">is_coalesced</span> <span class="o">=</span> <span class="n">data</span>
<span class="n">result</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">sparse_coo_tensor</span><span class="p">(</span>
<span class="n">indices</span><span class="p">,</span> <span class="n">values</span><span class="p">,</span> <span class="n">size</span><span class="p">,</span> <span class="n">check_invariants</span><span class="o">=</span><span class="kc">False</span><span class="p">,</span> <span class="n">is_coalesced</span><span class="o">=</span><span class="n">is_coalesced</span>
<span class="p">)</span>
<span class="n">_sparse_tensors_to_validate</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="n">result</span><span class="p">)</span>
<span class="k">return</span> <span class="n">result</span>
<span class="k">elif</span> <span class="n">layout</span> <span class="ow">in</span> <span class="p">{</span>
<span class="n">torch</span><span class="o">.</span><span class="n">sparse_csr</span><span class="p">,</span>
<span class="n">torch</span><span class="o">.</span><span class="n">sparse_csc</span><span class="p">,</span>
<span class="n">torch</span><span class="o">.</span><span class="n">sparse_bsr</span><span class="p">,</span>
<span class="n">torch</span><span class="o">.</span><span class="n">sparse_bsc</span><span class="p">,</span>
<span class="p">}:</span>
<span class="n">compressed_indices</span><span class="p">,</span> <span class="n">plain_indices</span><span class="p">,</span> <span class="n">values</span><span class="p">,</span> <span class="n">size</span> <span class="o">=</span> <span class="n">data</span>
<span class="n">result</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">sparse_compressed_tensor</span><span class="p">(</span>
<span class="n">compressed_indices</span><span class="p">,</span>
<span class="n">plain_indices</span><span class="p">,</span>
<span class="n">values</span><span class="p">,</span>
<span class="n">size</span><span class="p">,</span>
<span class="n">layout</span><span class="o">=</span><span class="n">layout</span><span class="p">,</span>
<span class="n">check_invariants</span><span class="o">=</span><span class="kc">False</span><span class="p">,</span>
<span class="p">)</span>
<span class="n">_sparse_tensors_to_validate</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="n">result</span><span class="p">)</span>
<span class="k">return</span> <span class="n">result</span>
<span class="k">raise</span> <span class="ne">NotImplementedError</span><span class="p">(</span><span class="sa">f</span><span class="s2">"rebuilding sparse tensor for layout </span><span class="si">{</span><span class="n">layout</span><span class="si">}</span><span class="s2">"</span><span class="p">)</span>
<span class="k">def</span> <span class="nf">_rebuild_nested_tensor</span><span class="p">(</span><span class="n">buffer</span><span class="p">,</span> <span class="n">sizes</span><span class="p">,</span> <span class="n">strides</span><span class="p">,</span> <span class="n">storage_offsets</span><span class="p">):</span>
<span class="k">return</span> <span class="n">torch</span><span class="o">.</span><span class="n">_nested_view_from_buffer</span><span class="p">(</span><span class="n">buffer</span><span class="p">,</span> <span class="n">sizes</span><span class="p">,</span> <span class="n">strides</span><span class="p">,</span> <span class="n">storage_offsets</span><span class="p">)</span>
<span class="k">def</span> <span class="nf">_rebuild_device_tensor_from_numpy</span><span class="p">(</span><span class="n">data</span><span class="p">,</span> <span class="n">dtype</span><span class="p">,</span> <span class="n">device</span><span class="p">,</span> <span class="n">requires_grad</span><span class="p">):</span>
<span class="n">tensor</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">from_numpy</span><span class="p">(</span><span class="n">data</span><span class="p">)</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="n">dtype</span><span class="o">=</span><span class="n">dtype</span><span class="p">,</span> <span class="n">device</span><span class="o">=</span><span class="n">device</span><span class="p">)</span>
<span class="n">tensor</span><span class="o">.</span><span class="n">requires_grad</span> <span class="o">=</span> <span class="n">requires_grad</span>
<span class="k">return</span> <span class="n">tensor</span>
<span class="c1"># Should not be used, only here to be able to load Tensors serialized with older versions of pytorch</span>
<span class="n">_rebuild_xla_tensor</span> <span class="o">=</span> <span class="n">_rebuild_device_tensor_from_numpy</span>
<span class="k">def</span> <span class="nf">_rebuild_meta_tensor_no_storage</span><span class="p">(</span><span class="n">dtype</span><span class="p">,</span> <span class="n">size</span><span class="p">,</span> <span class="n">stride</span><span class="p">,</span> <span class="n">requires_grad</span><span class="p">):</span>
<span class="k">return</span> <span class="n">torch</span><span class="o">.</span><span class="n">empty_strided</span><span class="p">(</span>
<span class="n">size</span><span class="p">,</span> <span class="n">stride</span><span class="p">,</span> <span class="n">dtype</span><span class="o">=</span><span class="n">dtype</span><span class="p">,</span> <span class="n">device</span><span class="o">=</span><span class="s2">"meta"</span><span class="p">,</span> <span class="n">requires_grad</span><span class="o">=</span><span class="n">requires_grad</span>
<span class="p">)</span>
<span class="k">def</span> <span class="nf">_rebuild_wrapper_subclass</span><span class="p">(</span>
<span class="bp">cls</span><span class="p">,</span> <span class="n">dtype</span><span class="p">,</span> <span class="n">size</span><span class="p">,</span> <span class="n">stride</span><span class="p">,</span> <span class="n">storage_offset</span><span class="p">,</span> <span class="n">layout</span><span class="p">,</span> <span class="n">device</span><span class="p">,</span> <span class="n">requires_grad</span>
<span class="p">):</span>
<span class="k">return</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="o">.</span><span class="n">_make_wrapper_subclass</span><span class="p">(</span> <span class="c1"># type: ignore[attr-defined]</span>
<span class="bp">cls</span><span class="p">,</span>
<span class="n">size</span><span class="p">,</span>
<span class="n">strides</span><span class="o">=</span><span class="n">stride</span><span class="p">,</span>
<span class="n">storage_offset</span><span class="o">=</span><span class="n">storage_offset</span><span class="p">,</span>
<span class="n">layout</span><span class="o">=</span><span class="n">layout</span><span class="p">,</span>
<span class="n">device</span><span class="o">=</span><span class="n">device</span><span class="p">,</span>
<span class="n">requires_grad</span><span class="o">=</span><span class="n">requires_grad</span><span class="p">,</span>
<span class="p">)</span>
<span class="c1"># TODO: Once we decide to break serialization FC, `storage` no longer needs to</span>
<span class="c1"># be a TypedStorage</span>
<span class="k">def</span> <span class="nf">_rebuild_qtensor</span><span class="p">(</span>
<span class="n">storage</span><span class="p">,</span>
<span class="n">storage_offset</span><span class="p">,</span>
<span class="n">size</span><span class="p">,</span>
<span class="n">stride</span><span class="p">,</span>
<span class="n">quantizer_params</span><span class="p">,</span>
<span class="n">requires_grad</span><span class="p">,</span>
<span class="n">backward_hooks</span><span class="p">,</span>
<span class="p">):</span>
<span class="n">qscheme</span> <span class="o">=</span> <span class="n">quantizer_params</span><span class="p">[</span><span class="mi">0</span><span class="p">]</span>
<span class="k">if</span> <span class="n">qscheme</span> <span class="o">==</span> <span class="n">torch</span><span class="o">.</span><span class="n">per_tensor_affine</span><span class="p">:</span>
<span class="n">_</span><span class="p">,</span> <span class="n">scale</span><span class="p">,</span> <span class="n">zero_point</span> <span class="o">=</span> <span class="n">quantizer_params</span>
<span class="n">tensor</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">_empty_affine_quantized</span><span class="p">(</span>
<span class="n">size</span><span class="p">,</span>
<span class="n">scale</span><span class="o">=</span><span class="n">scale</span><span class="p">,</span>
<span class="n">zero_point</span><span class="o">=</span><span class="n">zero_point</span><span class="p">,</span>
<span class="n">dtype</span><span class="o">=</span><span class="n">storage</span><span class="o">.</span><span class="n">dtype</span><span class="p">,</span>
<span class="n">device</span><span class="o">=</span><span class="n">storage</span><span class="o">.</span><span class="n">device</span><span class="p">,</span>
<span class="p">)</span>
<span class="k">elif</span> <span class="n">qscheme</span> <span class="ow">in</span> <span class="p">(</span><span class="n">torch</span><span class="o">.</span><span class="n">per_channel_affine</span><span class="p">,</span> <span class="n">torch</span><span class="o">.</span><span class="n">per_channel_affine_float_qparams</span><span class="p">):</span>
<span class="n">_</span><span class="p">,</span> <span class="n">scales</span><span class="p">,</span> <span class="n">zero_points</span><span class="p">,</span> <span class="n">axis</span> <span class="o">=</span> <span class="n">quantizer_params</span>
<span class="k">if</span> <span class="nb">type</span><span class="p">(</span><span class="n">scales</span><span class="p">)</span> <span class="ow">is</span> <span class="nb">list</span> <span class="ow">and</span> <span class="nb">type</span><span class="p">(</span><span class="n">zero_points</span><span class="p">)</span> <span class="ow">is</span> <span class="nb">list</span><span class="p">:</span>
<span class="k">if</span> <span class="n">qscheme</span> <span class="o">==</span> <span class="n">torch</span><span class="o">.</span><span class="n">per_channel_affine</span><span class="p">:</span>
<span class="n">scales</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">tensor</span><span class="p">(</span><span class="n">scales</span><span class="p">,</span> <span class="n">dtype</span><span class="o">=</span><span class="n">torch</span><span class="o">.</span><span class="n">double</span><span class="p">,</span> <span class="n">device</span><span class="o">=</span><span class="n">storage</span><span class="o">.</span><span class="n">device</span><span class="p">)</span>
<span class="n">zero_points</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">tensor</span><span class="p">(</span>
<span class="n">zero_points</span><span class="p">,</span> <span class="n">dtype</span><span class="o">=</span><span class="n">torch</span><span class="o">.</span><span class="n">long</span><span class="p">,</span> <span class="n">device</span><span class="o">=</span><span class="n">storage</span><span class="o">.</span><span class="n">device</span>
<span class="p">)</span>
<span class="k">else</span><span class="p">:</span>
<span class="n">scales</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">tensor</span><span class="p">(</span><span class="n">scales</span><span class="p">,</span> <span class="n">dtype</span><span class="o">=</span><span class="n">torch</span><span class="o">.</span><span class="n">float</span><span class="p">,</span> <span class="n">device</span><span class="o">=</span><span class="n">storage</span><span class="o">.</span><span class="n">device</span><span class="p">)</span>
<span class="n">zero_points</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">tensor</span><span class="p">(</span>
<span class="n">zero_points</span><span class="p">,</span> <span class="n">dtype</span><span class="o">=</span><span class="n">torch</span><span class="o">.</span><span class="n">float</span><span class="p">,</span> <span class="n">device</span><span class="o">=</span><span class="n">storage</span><span class="o">.</span><span class="n">device</span>
<span class="p">)</span>
<span class="n">tensor</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">_empty_per_channel_affine_quantized</span><span class="p">(</span>
<span class="n">size</span><span class="p">,</span>
<span class="n">scales</span><span class="o">=</span><span class="n">scales</span><span class="p">,</span>
<span class="n">zero_points</span><span class="o">=</span><span class="n">zero_points</span><span class="p">,</span>
<span class="n">axis</span><span class="o">=</span><span class="n">axis</span><span class="p">,</span>
<span class="n">dtype</span><span class="o">=</span><span class="n">storage</span><span class="o">.</span><span class="n">dtype</span><span class="p">,</span>
<span class="n">device</span><span class="o">=</span><span class="n">storage</span><span class="o">.</span><span class="n">device</span><span class="p">,</span>
<span class="p">)</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">"Can't deserialize quantized tensor with qscheme </span><span class="si">{</span><span class="n">qscheme</span><span class="si">}</span><span class="s2">"</span><span class="p">)</span>
<span class="n">tensor</span><span class="o">.</span><span class="n">set_</span><span class="p">(</span><span class="n">storage</span><span class="p">,</span> <span class="n">storage_offset</span><span class="p">,</span> <span class="n">size</span><span class="p">,</span> <span class="n">stride</span><span class="p">)</span>
<span class="n">tensor</span><span class="o">.</span><span class="n">requires_grad</span> <span class="o">=</span> <span class="n">requires_grad</span>
<span class="c1"># NB: This line exists only for backwards compatibility; the</span>
<span class="c1"># general expectation is that backward_hooks is an empty</span>
<span class="c1"># OrderedDict. See Note [Don't serialize hooks]</span>
<span class="n">tensor</span><span class="o">.</span><span class="n">_backward_hooks</span> <span class="o">=</span> <span class="n">backward_hooks</span>
<span class="k">return</span> <span class="n">tensor</span>
<span class="k">def</span> <span class="nf">_rebuild_parameter</span><span class="p">(</span><span class="n">data</span><span class="p">,</span> <span class="n">requires_grad</span><span class="p">,</span> <span class="n">backward_hooks</span><span class="p">):</span>
<span class="n">param</span> <span class="o">=</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="p">(</span><span class="n">data</span><span class="p">,</span> <span class="n">requires_grad</span><span class="p">)</span>
<span class="c1"># NB: This line exists only for backwards compatibility; the</span>
<span class="c1"># general expectation is that backward_hooks is an empty</span>
<span class="c1"># OrderedDict. See Note [Don't serialize hooks]</span>
<span class="n">param</span><span class="o">.</span><span class="n">_backward_hooks</span> <span class="o">=</span> <span class="n">backward_hooks</span>
<span class="k">return</span> <span class="n">param</span>
<span class="k">def</span> <span class="nf">_rebuild_parameter_with_state</span><span class="p">(</span><span class="n">data</span><span class="p">,</span> <span class="n">requires_grad</span><span class="p">,</span> <span class="n">backward_hooks</span><span class="p">,</span> <span class="n">state</span><span class="p">):</span>
<span class="n">param</span> <span class="o">=</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="p">(</span><span class="n">data</span><span class="p">,</span> <span class="n">requires_grad</span><span class="p">)</span>
<span class="c1"># NB: This line exists only for backwards compatibility; the</span>
<span class="c1"># general expectation is that backward_hooks is an empty</span>
<span class="c1"># OrderedDict. See Note [Don't serialize hooks]</span>
<span class="n">param</span><span class="o">.</span><span class="n">_backward_hooks</span> <span class="o">=</span> <span class="n">backward_hooks</span>
<span class="c1"># Restore state on Parameter like python attr.</span>
<span class="n">param</span> <span class="o">=</span> <span class="n">_set_obj_state</span><span class="p">(</span><span class="n">param</span><span class="p">,</span> <span class="n">state</span><span class="p">)</span>
<span class="k">return</span> <span class="n">param</span>
<span class="k">def</span> <span class="nf">_get_obj_state</span><span class="p">(</span><span class="n">obj</span><span class="p">):</span>
<span class="c1"># Get the state of the python subclass</span>
<span class="c1"># This loosely mimicks the function on the object class but since Tensor do not inherit</span>
<span class="c1"># from it, we cannot call that function directly</span>
<span class="c1"># https://github.com/python/cpython/blob/c83919bd635f4433f1c6ae8504996a9fe3c215e5/Objects/typeobject.c#L4891</span>
<span class="c1"># Note that starting with Python 3.11, this `__getstate__` is always defined and thus</span>
<span class="c1"># the else branch will never be taken.</span>
<span class="n">getstate_fn</span> <span class="o">=</span> <span class="nb">getattr</span><span class="p">(</span><span class="n">obj</span><span class="p">,</span> <span class="s2">"__getstate__"</span><span class="p">,</span> <span class="kc">None</span><span class="p">)</span>
<span class="k">if</span> <span class="n">getstate_fn</span><span class="p">:</span>
<span class="n">state</span> <span class="o">=</span> <span class="n">getstate_fn</span><span class="p">()</span>
<span class="k">else</span><span class="p">:</span>
<span class="n">slots_to_save</span> <span class="o">=</span> <span class="n">copyreg</span><span class="o">.</span><span class="n">_slotnames</span><span class="p">(</span><span class="n">obj</span><span class="o">.</span><span class="vm">__class__</span><span class="p">)</span> <span class="c1"># type: ignore[attr-defined]</span>
<span class="k">if</span> <span class="n">slots_to_save</span><span class="p">:</span>
<span class="n">state</span> <span class="o">=</span> <span class="p">(</span>
<span class="n">obj</span><span class="o">.</span><span class="vm">__dict__</span><span class="p">,</span>
<span class="p">{</span>
<span class="n">name</span><span class="p">:</span> <span class="nb">getattr</span><span class="p">(</span><span class="n">obj</span><span class="p">,</span> <span class="n">name</span><span class="p">)</span>
<span class="k">for</span> <span class="n">name</span> <span class="ow">in</span> <span class="n">slots_to_save</span>
<span class="k">if</span> <span class="nb">hasattr</span><span class="p">(</span><span class="n">obj</span><span class="p">,</span> <span class="n">name</span><span class="p">)</span>
<span class="p">},</span>
<span class="p">)</span>
<span class="k">else</span><span class="p">:</span>
<span class="n">state</span> <span class="o">=</span> <span class="n">obj</span><span class="o">.</span><span class="vm">__dict__</span>
<span class="k">return</span> <span class="n">state</span>
<span class="k">def</span> <span class="nf">_set_obj_state</span><span class="p">(</span><span class="n">obj</span><span class="p">,</span> <span class="n">state</span><span class="p">):</span>
<span class="k">if</span> <span class="nb">isinstance</span><span class="p">(</span><span class="n">state</span><span class="p">,</span> <span class="nb">tuple</span><span class="p">):</span>
<span class="k">if</span> <span class="ow">not</span> <span class="nb">len</span><span class="p">(</span><span class="n">state</span><span class="p">)</span> <span class="o">==</span> <span class="mi">2</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">"Invalid serialized state: </span><span class="si">{</span><span class="n">state</span><span class="si">}</span><span class="s2">"</span><span class="p">)</span>
<span class="n">dict_state</span> <span class="o">=</span> <span class="n">state</span><span class="p">[</span><span class="mi">0</span><span class="p">]</span>
<span class="n">slots_state</span> <span class="o">=</span> <span class="n">state</span><span class="p">[</span><span class="mi">1</span><span class="p">]</span>
<span class="k">else</span><span class="p">:</span>
<span class="n">dict_state</span> <span class="o">=</span> <span class="n">state</span>
<span class="n">slots_state</span> <span class="o">=</span> <span class="kc">None</span>
<span class="c1"># Starting with Python 3.11, the __dict__ attribute is lazily created</span>
<span class="c1"># and is serialized as None when not needed.</span>
<span class="k">if</span> <span class="n">dict_state</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">dict_state</span><span class="o">.</span><span class="n">items</span><span class="p">():</span>
<span class="nb">setattr</span><span class="p">(</span><span class="n">obj</span><span class="p">,</span> <span class="n">k</span><span class="p">,</span> <span class="n">v</span><span class="p">)</span>
<span class="k">if</span> <span class="n">slots_state</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">slots_state</span><span class="o">.</span><span class="n">items</span><span class="p">():</span>
<span class="nb">setattr</span><span class="p">(</span><span class="n">obj</span><span class="p">,</span> <span class="n">k</span><span class="p">,</span> <span class="n">v</span><span class="p">)</span>
<span class="k">return</span> <span class="n">obj</span>
<span class="k">def</span> <span class="nf">_import_dotted_name</span><span class="p">(</span><span class="n">name</span><span class="p">):</span>
<span class="n">components</span> <span class="o">=</span> <span class="n">name</span><span class="o">.</span><span class="n">split</span><span class="p">(</span><span class="s2">"."</span><span class="p">)</span>
<span class="n">obj</span> <span class="o">=</span> <span class="nb">__import__</span><span class="p">(</span><span class="n">components</span><span class="p">[</span><span class="mi">0</span><span class="p">])</span>
<span class="k">for</span> <span class="n">component</span> <span class="ow">in</span> <span class="n">components</span><span class="p">[</span><span class="mi">1</span><span class="p">:]:</span>
<span class="n">obj</span> <span class="o">=</span> <span class="nb">getattr</span><span class="p">(</span><span class="n">obj</span><span class="p">,</span> <span class="n">component</span><span class="p">)</span>
<span class="k">return</span> <span class="n">obj</span>
<span class="c1"># Taken from python 3.5 docs</span>
<span class="k">def</span> <span class="nf">_accumulate</span><span class="p">(</span><span class="n">iterable</span><span class="p">,</span> <span class="n">fn</span><span class="o">=</span><span class="k">lambda</span> <span class="n">x</span><span class="p">,</span> <span class="n">y</span><span class="p">:</span> <span class="n">x</span> <span class="o">+</span> <span class="n">y</span><span class="p">):</span>
<span class="s2">"Return running totals"</span>
<span class="c1"># _accumulate([1,2,3,4,5]) --> 1 3 6 10 15</span>
<span class="c1"># _accumulate([1,2,3,4,5], operator.mul) --> 1 2 6 24 120</span>
<span class="n">it</span> <span class="o">=</span> <span class="nb">iter</span><span class="p">(</span><span class="n">iterable</span><span class="p">)</span>
<span class="k">try</span><span class="p">:</span>
<span class="n">total</span> <span class="o">=</span> <span class="nb">next</span><span class="p">(</span><span class="n">it</span><span class="p">)</span>
<span class="k">except</span> <span class="ne">StopIteration</span><span class="p">:</span>
<span class="k">return</span>
<span class="k">yield</span> <span class="n">total</span>
<span class="k">for</span> <span class="n">element</span> <span class="ow">in</span> <span class="n">it</span><span class="p">:</span>
<span class="n">total</span> <span class="o">=</span> <span class="n">fn</span><span class="p">(</span><span class="n">total</span><span class="p">,</span> <span class="n">element</span><span class="p">)</span>
<span class="k">yield</span> <span class="n">total</span>
<span class="k">def</span> <span class="nf">_flatten_dense_tensors</span><span class="p">(</span><span class="n">tensors</span><span class="p">):</span>
<span class="w"> </span><span class="sd">"""Flatten dense tensors into a contiguous 1D buffer. Assume tensors are of</span>
<span class="sd"> same dense type.</span>
<span class="sd"> Since inputs are dense, the resulting tensor will be a concatenated 1D</span>
<span class="sd"> buffer. Element-wise operation on this buffer will be equivalent to</span>
<span class="sd"> operating individually.</span>
<span class="sd"> Args:</span>
<span class="sd"> tensors (Iterable[Tensor]): dense tensors to flatten.</span>
<span class="sd"> Returns:</span>
<span class="sd"> A contiguous 1D buffer containing input tensors.</span>
<span class="sd"> """</span>
<span class="k">return</span> <span class="n">torch</span><span class="o">.</span><span class="n">_C</span><span class="o">.</span><span class="n">_nn</span><span class="o">.</span><span class="n">flatten_dense_tensors</span><span class="p">(</span><span class="n">tensors</span><span class="p">)</span>
<span class="k">def</span> <span class="nf">_flatten_sparse_tensors</span><span class="p">(</span><span class="n">tensors</span><span class="p">):</span>
<span class="w"> </span><span class="sd">"""Flatten sparse tensors into two contiguous 1D buffers, one of indices and</span>
<span class="sd"> one of values. Assume tensors are of same sparse type.</span>
<span class="sd"> Args:</span>
<span class="sd"> tensors (Iterable[Tensor]): sparse tensors to flatten.</span>
<span class="sd"> Returns:</span>
<span class="sd"> A tuple of two contiguous 1D buffers, one containing input tensors'</span>
<span class="sd"> indices and the other containing the values.</span>