-
Notifications
You must be signed in to change notification settings - Fork 2
/
Copy pathnested.html
1283 lines (1100 loc) · 93.6 KB
/
nested.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.nested — PyTorch 2.5 documentation</title>
<link rel="canonical" href="https://pytorch.org/docs/stable/nested.html"/>
<link rel="stylesheet" href="_static/css/theme.css" type="text/css" />
<!-- <link rel="stylesheet" href="_static/pygments.css" type="text/css" /> -->
<link rel="stylesheet" href="_static/pygments.css" type="text/css" />
<link rel="stylesheet" href="_static/css/theme.css" type="text/css" />
<link rel="stylesheet" href="_static/copybutton.css" type="text/css" />
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/[email protected]/dist/katex.min.css" type="text/css" />
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/[email protected]/dist/katex.min.css" type="text/css" />
<link rel="stylesheet" href="_static/katex-math.css" type="text/css" />
<link rel="stylesheet" href="_static/sphinx-dropdown.css" type="text/css" />
<link rel="stylesheet" href="_static/panels-bootstrap.min.css" type="text/css" />
<link rel="stylesheet" href="_static/css/jit.css" type="text/css" />
<link rel="index" title="Index" href="genindex.html" />
<link rel="search" title="Search" href="search.html" />
<link rel="next" title="torch.Size" href="size.html" />
<link rel="prev" title="torch.masked" href="masked.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 class="main-menu-item">
<div id="resourcesDropdownButton" data-toggle="resources-dropdown" class="resources-dropdown">
<a class="with-down-arrow">
Learn
</a>
<div class="resources-dropdown-menu">
<a class="nav-dropdown-item" href="https://pytorch.org/get-started">
<span class=dropdown-title>Get Started</span>
<p>Run PyTorch locally or get started quickly with one of the supported cloud platforms</p>
</a>
<a class="nav-dropdown-item" href="https://pytorch.org/tutorials">
<span class="dropdown-title">Tutorials</span>
<p>Whats new in PyTorch tutorials</p>
</a>
<a class="nav-dropdown-item" href="https://pytorch.org/tutorials/beginner/basics/intro.html">
<span class="dropdown-title">Learn the Basics</span>
<p>Familiarize yourself with PyTorch concepts and modules</p>
</a>
<a class="nav-dropdown-item" href="https://pytorch.org/tutorials/recipes/recipes_index.html">
<span class="dropdown-title">PyTorch Recipes</span>
<p>Bite-size, ready-to-deploy PyTorch code examples</p>
</a>
<a class="nav-dropdown-item" href="https://pytorch.org/tutorials/beginner/introyt.html">
<span class="dropdown-title">Intro to PyTorch - YouTube Series</span>
<p>Master PyTorch basics with our engaging YouTube tutorial series</p>
</a>
</div>
</div>
</li>
<li>
<div id="resourcesDropdownButton" data-toggle="resources-dropdown" class="resources-dropdown">
<a class="with-down-arrow">
Ecosystem
</a>
<div class="resources-dropdown-menu">
<a class="nav-dropdown-item" href="https://pytorch.org/ecosystem">
<span class="dropdown-title">Tools</span>
<p>Learn about the tools and frameworks in the PyTorch Ecosystem</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://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/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/ecosystem/contributor-awards-2023">
<span class="dropdown-title">Contributor Awards - 2023</span>
<p>Award winners announced at this year's PyTorch Conference</p>
</a>
</div>
</div>
</li>
<li>
<div id="resourcesDropdownButton" data-toggle="resources-dropdown" class="resources-dropdown">
<a class="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>
<p>Build innovative and privacy-aware AI experiences for edge devices</p>
</a>
<a class="nav-dropdown-item" href="https://pytorch.org/executorch-overview">
<span class="dropdown-title">ExecuTorch</span>
<p>End-to-end solution for enabling on-device inference capabilities across mobile and edge devices</p>
</a>
</div>
</div>
</li>
<li class="main-menu-item">
<div id="resourcesDropdownButton" data-toggle="resources-dropdown" class="resources-dropdown">
<a class="with-down-arrow">
Docs
</a>
<div class="resources-dropdown-menu">
<a class="nav-dropdown-item" href="https://pytorch.org/docs/stable/index.html">
<span class="dropdown-title">PyTorch</span>
<p>Explore the documentation for comprehensive guidance on how to use PyTorch</p>
</a>
<a class="nav-dropdown-item" href="https://pytorch.org/pytorch-domains">
<span class="dropdown-title">PyTorch Domains</span>
<p>Read the PyTorch Domains documentation to learn more about domain-specific libraries</p>
</a>
</div>
</div>
</li>
<li>
<div id="resourcesDropdownButton" data-toggle="resources-dropdown" class="resources-dropdown">
<a class="with-down-arrow">
Blogs & News
</a>
<div class="resources-dropdown-menu">
<a class="nav-dropdown-item" href="https://pytorch.org/blog/">
<span class="dropdown-title">PyTorch Blog</span>
<p>Catch up on the latest technical news and happenings</p>
</a>
<a class="nav-dropdown-item" href="https://pytorch.org/community-blog">
<span class="dropdown-title">Community Blog</span>
<p>Stories from the PyTorch ecosystem</p>
</a>
<a class="nav-dropdown-item" href="https://pytorch.org/videos">
<span class="dropdown-title">Videos</span>
<p>Learn about the latest PyTorch tutorials, new, and more </p>
<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/events">
<span class="dropdown-title">Events</span>
<p>Find events, webinars, and podcasts</p>
</a>
</div>
</li>
<li>
<div id="resourcesDropdownButton" data-toggle="resources-dropdown" class="resources-dropdown">
<a class="with-down-arrow">
About
</a>
<div class="resources-dropdown-menu">
<a class="nav-dropdown-item" href="https://pytorch.org/foundation">
<span class="dropdown-title">PyTorch Foundation</span>
<p>Learn more about the PyTorch Foundation</p>
</a>
<a class="nav-dropdown-item" href="https://pytorch.org/governing-board">
<span class="dropdown-title">Governing Board</span>
<p></p>
</a>
</div>
</div>
</li>
<li class="main-menu-item">
<div class="no-dropdown">
<a href="https://pytorch.org/join" data-cta="join">
Become a Member
</a>
</div>
</li>
<li>
<div class="main-menu-item">
<a href="https://github.com/pytorch/pytorch" class="github-icon">
</a>
</div>
</li>
<!--- TODO: This block adds the search icon to the nav bar. We will enable it later.
<li>
<div class="main-menu-item">
<a href="https://github.com/pytorch/pytorch" class="search-icon">
</a>
</div>
</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'>2.5 ▼</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">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/custom_operators.html">PyTorch Custom Operators Landing Page</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/fsdp.html">FSDP Notes</a></li>
<li class="toctree-l1"><a class="reference internal" href="notes/get_start_xpu.html">Getting Started on Intel GPU</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 class="current">
<li class="toctree-l1"><a class="reference internal" href="torch.html">torch</a></li>
<li class="toctree-l1"><a class="reference internal" href="nn.html">torch.nn</a></li>
<li class="toctree-l1"><a class="reference internal" href="nn.functional.html">torch.nn.functional</a></li>
<li class="toctree-l1"><a class="reference internal" href="tensors.html">torch.Tensor</a></li>
<li class="toctree-l1"><a class="reference internal" href="tensor_attributes.html">Tensor Attributes</a></li>
<li class="toctree-l1"><a class="reference internal" href="tensor_view.html">Tensor Views</a></li>
<li class="toctree-l1"><a class="reference internal" href="amp.html">torch.amp</a></li>
<li class="toctree-l1"><a class="reference internal" href="autograd.html">torch.autograd</a></li>
<li class="toctree-l1"><a class="reference internal" href="library.html">torch.library</a></li>
<li class="toctree-l1"><a class="reference internal" href="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="xpu.html">torch.xpu</a></li>
<li class="toctree-l1"><a class="reference internal" href="mtia.html">torch.mtia</a></li>
<li class="toctree-l1"><a class="reference internal" href="meta.html">Meta device</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.tensor.html">torch.distributed.tensor</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.tensor.parallel.html">torch.distributed.tensor.parallel</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.pipelining.html">torch.distributed.pipelining</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="fx.experimental.html">torch.fx.experimental</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="nn.attention.html">torch.nn.attention</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="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 current"><a class="current reference internal" href="#">torch.nested</a></li>
<li class="toctree-l1"><a class="reference internal" href="size.html">torch.Size</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="module_tracker.html">torch.utils.module_tracker</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="future_mod.html">torch.__future__</a></li>
<li class="toctree-l1"><a class="reference internal" href="logging.html">torch._logging</a></li>
<li class="toctree-l1"><a class="reference internal" href="torch_environment_variables.html">Torch Environment Variables</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>torch.nested</li>
<li class="pytorch-breadcrumbs-aside">
<a href="_sources/nested.rst.txt" rel="nofollow"><img src="_static/images/view-page-source-icon.svg"></a>
</li>
</ul>
</div>
</div>
<div class="pytorch-shortcuts-wrapper" id="pytorch-shortcuts-wrapper">
Shortcuts
</div>
</div>
<section data-toggle="wy-nav-shift" id="pytorch-content-wrap" class="pytorch-content-wrap">
<div class="pytorch-content-left">
<!-- 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">
<div class="section" id="module-torch.nested">
<span id="torch-nested"></span><h1>torch.nested<a class="headerlink" href="#module-torch.nested" title="Permalink to this heading">¶</a></h1>
<div class="section" id="introduction">
<h2>Introduction<a class="headerlink" href="#introduction" title="Permalink to this heading">¶</a></h2>
<div class="admonition warning">
<p class="admonition-title">Warning</p>
<p>The PyTorch API of nested tensors is in prototype stage and will change in the near future.</p>
</div>
<p>NestedTensor allows the user to pack a list of Tensors into a single, efficient datastructure.</p>
<p>The only constraint on the input Tensors is that their dimension must match.</p>
<p>This enables more efficient metadata representations and access to purpose built kernels.</p>
<p>One application of NestedTensors is to express sequential data in various domains.
While the conventional approach is to pad variable length sequences, NestedTensor
enables users to bypass padding. The API for calling operations on a nested tensor is no different
from that of a regular <code class="docutils literal notranslate"><span class="pre">torch.Tensor</span></code>, which should allow seamless integration with existing models,
with the main difference being <a class="reference internal" href="#construction"><span class="std std-ref">construction of the inputs</span></a>.</p>
<p>As this is a prototype feature, the <a class="reference internal" href="#supported-operations"><span class="std std-ref">operations supported</span></a> are still
limited. However, we welcome issues, feature requests and contributions. More information on contributing can be found
<a class="reference external" href="https://github.com/pytorch/pytorch/blob/main/aten/src/ATen/native/nested/README.md">in this Readme</a>.</p>
</div>
<div class="section" id="construction">
<span id="id1"></span><h2>Construction<a class="headerlink" href="#construction" title="Permalink to this heading">¶</a></h2>
<p>Construction is straightforward and involves passing a list of Tensors to the <code class="docutils literal notranslate"><span class="pre">torch.nested.nested_tensor</span></code>
constructor.</p>
<div class="doctest highlight-default notranslate"><div class="highlight"><pre><span></span><span class="gp">>>> </span><span class="n">a</span><span class="p">,</span> <span class="n">b</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">arange</span><span class="p">(</span><span class="mi">3</span><span class="p">),</span> <span class="n">torch</span><span class="o">.</span><span class="n">arange</span><span class="p">(</span><span class="mi">5</span><span class="p">)</span> <span class="o">+</span> <span class="mi">3</span>
<span class="gp">>>> </span><span class="n">a</span>
<span class="go">tensor([0, 1, 2])</span>
<span class="gp">>>> </span><span class="n">b</span>
<span class="go">tensor([3, 4, 5, 6, 7])</span>
<span class="gp">>>> </span><span class="n">nt</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">nested</span><span class="o">.</span><span class="n">nested_tensor</span><span class="p">([</span><span class="n">a</span><span class="p">,</span> <span class="n">b</span><span class="p">])</span>
<span class="gp">>>> </span><span class="n">nt</span>
<span class="go">nested_tensor([</span>
<span class="go"> tensor([0, 1, 2]),</span>
<span class="go"> tensor([3, 4, 5, 6, 7])</span>
<span class="go"> ])</span>
</pre></div>
</div>
<p>Data type, device and whether gradients are required can be chosen via the usual keyword arguments.</p>
<div class="doctest highlight-default notranslate"><div class="highlight"><pre><span></span><span class="gp">>>> </span><span class="n">nt</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">nested</span><span class="o">.</span><span class="n">nested_tensor</span><span class="p">([</span><span class="n">a</span><span class="p">,</span> <span class="n">b</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">float32</span><span class="p">,</span> <span class="n">device</span><span class="o">=</span><span class="s2">"cuda"</span><span class="p">,</span> <span class="n">requires_grad</span><span class="o">=</span><span class="kc">True</span><span class="p">)</span>
<span class="gp">>>> </span><span class="n">nt</span>
<span class="go">nested_tensor([</span>
<span class="go"> tensor([0., 1., 2.], device='cuda:0', requires_grad=True),</span>
<span class="go"> tensor([3., 4., 5., 6., 7.], device='cuda:0', requires_grad=True)</span>
<span class="go">], device='cuda:0', requires_grad=True)</span>
</pre></div>
</div>
<p>In the vein of <code class="docutils literal notranslate"><span class="pre">torch.as_tensor</span></code>, <code class="docutils literal notranslate"><span class="pre">torch.nested.as_nested_tensor</span></code> can be used to preserve autograd
history from the tensors passed to the constructor. For more information, refer to the section on
<a class="reference internal" href="#constructor-functions"><span class="std std-ref">Nested tensor constructor and conversion functions</span></a>.</p>
<p>In order to form a valid NestedTensor all the passed Tensors need to match in dimension, but none of the other attributes need to.</p>
<div class="doctest highlight-default notranslate"><div class="highlight"><pre><span></span><span class="gp">>>> </span><span class="n">a</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">randn</span><span class="p">(</span><span class="mi">3</span><span class="p">,</span> <span class="mi">50</span><span class="p">,</span> <span class="mi">70</span><span class="p">)</span> <span class="c1"># image 1</span>
<span class="gp">>>> </span><span class="n">b</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">randn</span><span class="p">(</span><span class="mi">3</span><span class="p">,</span> <span class="mi">128</span><span class="p">,</span> <span class="mi">64</span><span class="p">)</span> <span class="c1"># image 2</span>
<span class="gp">>>> </span><span class="n">nt</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">nested</span><span class="o">.</span><span class="n">nested_tensor</span><span class="p">([</span><span class="n">a</span><span class="p">,</span> <span class="n">b</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">float32</span><span class="p">)</span>
<span class="gp">>>> </span><span class="n">nt</span><span class="o">.</span><span class="n">dim</span><span class="p">()</span>
<span class="go">4</span>
</pre></div>
</div>
<p>If one of the dimensions doesn’t match, the constructor throws an error.</p>
<div class="doctest highlight-default notranslate"><div class="highlight"><pre><span></span><span class="gp">>>> </span><span class="n">a</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">randn</span><span class="p">(</span><span class="mi">50</span><span class="p">,</span> <span class="mi">128</span><span class="p">)</span> <span class="c1"># text 1</span>
<span class="gp">>>> </span><span class="n">b</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">randn</span><span class="p">(</span><span class="mi">3</span><span class="p">,</span> <span class="mi">128</span><span class="p">,</span> <span class="mi">64</span><span class="p">)</span> <span class="c1"># image 2</span>
<span class="gp">>>> </span><span class="n">nt</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">nested</span><span class="o">.</span><span class="n">nested_tensor</span><span class="p">([</span><span class="n">a</span><span class="p">,</span> <span class="n">b</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">float32</span><span class="p">)</span>
<span class="gt">Traceback (most recent call last):</span>
File <span class="nb">"<stdin>"</span>, line <span class="m">1</span>, in <span class="n"><module></span>
<span class="gr">RuntimeError</span>: <span class="n">All Tensors given to nested_tensor must have the same dimension. Found dimension 3 for Tensor at index 1 and dimension 2 for Tensor at index 0.</span>
</pre></div>
</div>
<p>Note that the passed Tensors are being copied into a contiguous piece of memory. The resulting
NestedTensor allocates new memory to store them and does not keep a reference.</p>
<p>At this moment we only support one level of nesting, i.e. a simple, flat list of Tensors. In the future
we can add support for multiple levels of nesting, such as a list that consists entirely of lists of Tensors.
Note that for this extension it is important to maintain an even level of nesting across entries so that the resulting NestedTensor
has a well defined dimension. If you have a need for this feature, please feel encouraged to open a feature request so that
we can track it and plan accordingly.</p>
</div>
<div class="section" id="size">
<h2>size<a class="headerlink" href="#size" title="Permalink to this heading">¶</a></h2>
<p>Even though a NestedTensor does not support <code class="docutils literal notranslate"><span class="pre">.size()</span></code> (or <code class="docutils literal notranslate"><span class="pre">.shape</span></code>), it supports <code class="docutils literal notranslate"><span class="pre">.size(i)</span></code> if dimension i is regular.</p>
<div class="doctest highlight-default notranslate"><div class="highlight"><pre><span></span><span class="gp">>>> </span><span class="n">a</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">randn</span><span class="p">(</span><span class="mi">50</span><span class="p">,</span> <span class="mi">128</span><span class="p">)</span> <span class="c1"># text 1</span>
<span class="gp">>>> </span><span class="n">b</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">randn</span><span class="p">(</span><span class="mi">32</span><span class="p">,</span> <span class="mi">128</span><span class="p">)</span> <span class="c1"># text 2</span>
<span class="gp">>>> </span><span class="n">nt</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">nested</span><span class="o">.</span><span class="n">nested_tensor</span><span class="p">([</span><span class="n">a</span><span class="p">,</span> <span class="n">b</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">float32</span><span class="p">)</span>
<span class="gp">>>> </span><span class="n">nt</span><span class="o">.</span><span class="n">size</span><span class="p">(</span><span class="mi">0</span><span class="p">)</span>
<span class="go">2</span>
<span class="gp">>>> </span><span class="n">nt</span><span class="o">.</span><span class="n">size</span><span class="p">(</span><span class="mi">1</span><span class="p">)</span>
<span class="gt">Traceback (most recent call last):</span>
File <span class="nb">"<stdin>"</span>, line <span class="m">1</span>, in <span class="n"><module></span>
<span class="gr">RuntimeError</span>: <span class="n">Given dimension 1 is irregular and does not have a size.</span>
<span class="gp">>>> </span><span class="n">nt</span><span class="o">.</span><span class="n">size</span><span class="p">(</span><span class="mi">2</span><span class="p">)</span>
<span class="go">128</span>
</pre></div>
</div>
<p>If all dimensions are regular, the NestedTensor is intended to be semantically indistinguishable from a regular <code class="docutils literal notranslate"><span class="pre">torch.Tensor</span></code>.</p>
<div class="doctest highlight-default notranslate"><div class="highlight"><pre><span></span><span class="gp">>>> </span><span class="n">a</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">randn</span><span class="p">(</span><span class="mi">20</span><span class="p">,</span> <span class="mi">128</span><span class="p">)</span> <span class="c1"># text 1</span>
<span class="gp">>>> </span><span class="n">nt</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">nested</span><span class="o">.</span><span class="n">nested_tensor</span><span class="p">([</span><span class="n">a</span><span class="p">,</span> <span class="n">a</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">float32</span><span class="p">)</span>
<span class="gp">>>> </span><span class="n">nt</span><span class="o">.</span><span class="n">size</span><span class="p">(</span><span class="mi">0</span><span class="p">)</span>
<span class="go">2</span>
<span class="gp">>>> </span><span class="n">nt</span><span class="o">.</span><span class="n">size</span><span class="p">(</span><span class="mi">1</span><span class="p">)</span>
<span class="go">20</span>
<span class="gp">>>> </span><span class="n">nt</span><span class="o">.</span><span class="n">size</span><span class="p">(</span><span class="mi">2</span><span class="p">)</span>
<span class="go">128</span>
<span class="gp">>>> </span><span class="n">torch</span><span class="o">.</span><span class="n">stack</span><span class="p">(</span><span class="n">nt</span><span class="o">.</span><span class="n">unbind</span><span class="p">())</span><span class="o">.</span><span class="n">size</span><span class="p">()</span>
<span class="go">torch.Size([2, 20, 128])</span>
<span class="gp">>>> </span><span class="n">torch</span><span class="o">.</span><span class="n">stack</span><span class="p">([</span><span class="n">a</span><span class="p">,</span> <span class="n">a</span><span class="p">])</span><span class="o">.</span><span class="n">size</span><span class="p">()</span>
<span class="go">torch.Size([2, 20, 128])</span>
<span class="gp">>>> </span><span class="n">torch</span><span class="o">.</span><span class="n">equal</span><span class="p">(</span><span class="n">torch</span><span class="o">.</span><span class="n">stack</span><span class="p">(</span><span class="n">nt</span><span class="o">.</span><span class="n">unbind</span><span class="p">()),</span> <span class="n">torch</span><span class="o">.</span><span class="n">stack</span><span class="p">([</span><span class="n">a</span><span class="p">,</span> <span class="n">a</span><span class="p">]))</span>
<span class="go">True</span>
</pre></div>
</div>
<p>In the future we might make it easier to detect this condition and convert seamlessly.</p>
<p>Please open a feature request if you have a need for this (or any other related feature for that matter).</p>
</div>
<div class="section" id="unbind">
<h2>unbind<a class="headerlink" href="#unbind" title="Permalink to this heading">¶</a></h2>
<p><code class="docutils literal notranslate"><span class="pre">unbind</span></code> allows you to retrieve a view of the constituents.</p>
<div class="doctest highlight-default notranslate"><div class="highlight"><pre><span></span><span class="gp">>>> </span><span class="kn">import</span> <span class="nn">torch</span>
<span class="gp">>>> </span><span class="n">a</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">randn</span><span class="p">(</span><span class="mi">2</span><span class="p">,</span> <span class="mi">3</span><span class="p">)</span>
<span class="gp">>>> </span><span class="n">b</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">randn</span><span class="p">(</span><span class="mi">3</span><span class="p">,</span> <span class="mi">4</span><span class="p">)</span>
<span class="gp">>>> </span><span class="n">nt</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">nested</span><span class="o">.</span><span class="n">nested_tensor</span><span class="p">([</span><span class="n">a</span><span class="p">,</span> <span class="n">b</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">float32</span><span class="p">)</span>
<span class="gp">>>> </span><span class="n">nt</span>
<span class="go">nested_tensor([</span>
<span class="go"> tensor([[ 1.2286, -1.2343, -1.4842],</span>
<span class="go"> [-0.7827, 0.6745, 0.0658]]),</span>
<span class="go"> tensor([[-1.1247, -0.4078, -1.0633, 0.8083],</span>
<span class="go"> [-0.2871, -0.2980, 0.5559, 1.9885],</span>
<span class="go"> [ 0.4074, 2.4855, 0.0733, 0.8285]])</span>
<span class="go">])</span>
<span class="gp">>>> </span><span class="n">nt</span><span class="o">.</span><span class="n">unbind</span><span class="p">()</span>
<span class="go">(tensor([[ 1.2286, -1.2343, -1.4842],</span>
<span class="go"> [-0.7827, 0.6745, 0.0658]]), tensor([[-1.1247, -0.4078, -1.0633, 0.8083],</span>
<span class="go"> [-0.2871, -0.2980, 0.5559, 1.9885],</span>
<span class="go"> [ 0.4074, 2.4855, 0.0733, 0.8285]]))</span>
<span class="gp">>>> </span><span class="n">nt</span><span class="o">.</span><span class="n">unbind</span><span class="p">()[</span><span class="mi">0</span><span class="p">]</span> <span class="ow">is</span> <span class="ow">not</span> <span class="n">a</span>
<span class="go">True</span>
<span class="gp">>>> </span><span class="n">nt</span><span class="o">.</span><span class="n">unbind</span><span class="p">()[</span><span class="mi">0</span><span class="p">]</span><span class="o">.</span><span class="n">mul_</span><span class="p">(</span><span class="mi">3</span><span class="p">)</span>
<span class="go">tensor([[ 3.6858, -3.7030, -4.4525],</span>
<span class="go"> [-2.3481, 2.0236, 0.1975]])</span>
<span class="gp">>>> </span><span class="n">nt</span>
<span class="go">nested_tensor([</span>
<span class="go"> tensor([[ 3.6858, -3.7030, -4.4525],</span>
<span class="go"> [-2.3481, 2.0236, 0.1975]]),</span>
<span class="go"> tensor([[-1.1247, -0.4078, -1.0633, 0.8083],</span>
<span class="go"> [-0.2871, -0.2980, 0.5559, 1.9885],</span>
<span class="go"> [ 0.4074, 2.4855, 0.0733, 0.8285]])</span>
<span class="go">])</span>
</pre></div>
</div>
<p>Note that <code class="docutils literal notranslate"><span class="pre">nt.unbind()[0]</span></code> is not a copy, but rather a slice of the underlying memory, which represents the first entry or constituent of the NestedTensor.</p>
</div>
<div class="section" id="nested-tensor-constructor-and-conversion-functions">
<span id="constructor-functions"></span><h2>Nested tensor constructor and conversion functions<a class="headerlink" href="#nested-tensor-constructor-and-conversion-functions" title="Permalink to this heading">¶</a></h2>
<p>The following functions are related to nested tensors:</p>
<dl class="py function">
<dt class="sig sig-object py" id="torch.nested.nested_tensor">
<span class="sig-prename descclassname"><span class="pre">torch.nested.</span></span><span class="sig-name descname"><span class="pre">nested_tensor</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">tensor_list</span></span></em>, <em class="sig-param"><span class="o"><span class="pre">*</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">dtype</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">layout</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">device</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">requires_grad</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">False</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">pin_memory</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">False</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/torch/nested.html#nested_tensor"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#torch.nested.nested_tensor" title="Permalink to this definition">¶</a></dt>
<dd><p>Constructs a nested tensor with no autograd history (also known as a “leaf tensor”, see
<a class="reference internal" href="notes/autograd.html#autograd-mechanics"><span class="std std-ref">Autograd mechanics</span></a>) from <code class="xref py py-attr docutils literal notranslate"><span class="pre">tensor_list</span></code> a list of tensors.</p>
<dl class="field-list simple">
<dt class="field-odd">Parameters</dt>
<dd class="field-odd"><ul class="simple">
<li><p><strong>tensor_list</strong> (<em>List</em><em>[</em><em>array_like</em><em>]</em>) – a list of tensors, or anything that can be passed to torch.tensor,</p></li>
<li><p><strong>dimensionality.</strong> (<em>where each element</em><em> of </em><em>the list has the same</em>) – </p></li>
</ul>
</dd>
<dt class="field-even">Keyword Arguments</dt>
<dd class="field-even"><ul class="simple">
<li><p><strong>dtype</strong> (<a class="reference internal" href="tensor_attributes.html#torch.dtype" title="torch.dtype"><code class="xref py py-class docutils literal notranslate"><span class="pre">torch.dtype</span></code></a>, optional) – the desired type of returned nested tensor.
Default: if None, same <a class="reference internal" href="tensor_attributes.html#torch.dtype" title="torch.dtype"><code class="xref py py-class docutils literal notranslate"><span class="pre">torch.dtype</span></code></a> as leftmost tensor in the list.</p></li>
<li><p><strong>layout</strong> (<a class="reference internal" href="tensor_attributes.html#torch.layout" title="torch.layout"><code class="xref py py-class docutils literal notranslate"><span class="pre">torch.layout</span></code></a>, optional) – the desired layout of returned nested tensor.
Only strided and jagged layouts are supported. Default: if None, the strided layout.</p></li>
<li><p><strong>device</strong> (<a class="reference internal" href="tensor_attributes.html#torch.device" title="torch.device"><code class="xref py py-class docutils literal notranslate"><span class="pre">torch.device</span></code></a>, optional) – the desired device of returned nested tensor.
Default: if None, same <a class="reference internal" href="tensor_attributes.html#torch.device" title="torch.device"><code class="xref py py-class docutils literal notranslate"><span class="pre">torch.device</span></code></a> as leftmost tensor in the list</p></li>
<li><p><strong>requires_grad</strong> (<a class="reference external" href="https://docs.python.org/3/library/functions.html#bool" title="(in Python v3.13)"><em>bool</em></a><em>, </em><em>optional</em>) – If autograd should record operations on the
returned nested tensor. Default: <code class="docutils literal notranslate"><span class="pre">False</span></code>.</p></li>
<li><p><strong>pin_memory</strong> (<a class="reference external" href="https://docs.python.org/3/library/functions.html#bool" title="(in Python v3.13)"><em>bool</em></a><em>, </em><em>optional</em>) – If set, returned nested tensor would be allocated in
the pinned memory. Works only for CPU tensors. Default: <code class="docutils literal notranslate"><span class="pre">False</span></code>.</p></li>
</ul>
</dd>
<dt class="field-odd">Return type</dt>
<dd class="field-odd"><p><a class="reference internal" href="tensors.html#torch.Tensor" title="torch.Tensor"><em>Tensor</em></a></p>
</dd>
</dl>
<p>Example:</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="gp">>>> </span><span class="n">a</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">arange</span><span class="p">(</span><span class="mi">3</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">requires_grad</span><span class="o">=</span><span class="kc">True</span><span class="p">)</span>
<span class="gp">>>> </span><span class="n">b</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">arange</span><span class="p">(</span><span class="mi">5</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">requires_grad</span><span class="o">=</span><span class="kc">True</span><span class="p">)</span>
<span class="gp">>>> </span><span class="n">nt</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">nested</span><span class="o">.</span><span class="n">nested_tensor</span><span class="p">([</span><span class="n">a</span><span class="p">,</span> <span class="n">b</span><span class="p">],</span> <span class="n">requires_grad</span><span class="o">=</span><span class="kc">True</span><span class="p">)</span>
<span class="gp">>>> </span><span class="n">nt</span><span class="o">.</span><span class="n">is_leaf</span>
<span class="go">True</span>
</pre></div>
</div>
</dd></dl>
<dl class="py function">
<dt class="sig sig-object py" id="torch.nested.as_nested_tensor">
<span class="sig-prename descclassname"><span class="pre">torch.nested.</span></span><span class="sig-name descname"><span class="pre">as_nested_tensor</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">ts</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">dtype</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">device</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">layout</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/torch/nested.html#as_nested_tensor"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#torch.nested.as_nested_tensor" title="Permalink to this definition">¶</a></dt>
<dd><p>Constructs a nested tensor preserving autograd history from a tensor or a list / tuple of
tensors.</p>
<p>If a nested tensor is passed, it will be returned directly unless the device / dtype / layout
differ. Note that converting device / dtype will result in a copy, while converting layout
is not currently supported by this function.</p>
<p>If a non-nested tensor is passed, it is treated as a batch of constituents of consistent size.
A copy will be incurred if the passed device / dtype differ from those of the input OR if
the input is non-contiguous. Otherwise, the input’s storage will be used directly.</p>
<p>If a tensor list is provided, tensors in the list are always copied during construction of
the nested tensor.</p>
<dl class="field-list simple">
<dt class="field-odd">Parameters</dt>
<dd class="field-odd"><p><strong>ts</strong> (<a class="reference internal" href="tensors.html#torch.Tensor" title="torch.Tensor"><em>Tensor</em></a><em> or </em><em>List</em><em>[</em><a class="reference internal" href="tensors.html#torch.Tensor" title="torch.Tensor"><em>Tensor</em></a><em>] or </em><em>Tuple</em><em>[</em><a class="reference internal" href="tensors.html#torch.Tensor" title="torch.Tensor"><em>Tensor</em></a><em>]</em>) – a tensor to treat as a nested tensor OR a
list / tuple of tensors with the same ndim</p>
</dd>
<dt class="field-even">Keyword Arguments</dt>
<dd class="field-even"><ul class="simple">
<li><p><strong>dtype</strong> (<a class="reference internal" href="tensor_attributes.html#torch.dtype" title="torch.dtype"><code class="xref py py-class docutils literal notranslate"><span class="pre">torch.dtype</span></code></a>, optional) – the desired type of returned nested tensor.
Default: if None, same <a class="reference internal" href="tensor_attributes.html#torch.dtype" title="torch.dtype"><code class="xref py py-class docutils literal notranslate"><span class="pre">torch.dtype</span></code></a> as leftmost tensor in the list.</p></li>
<li><p><strong>device</strong> (<a class="reference internal" href="tensor_attributes.html#torch.device" title="torch.device"><code class="xref py py-class docutils literal notranslate"><span class="pre">torch.device</span></code></a>, optional) – the desired device of returned nested tensor.
Default: if None, same <a class="reference internal" href="tensor_attributes.html#torch.device" title="torch.device"><code class="xref py py-class docutils literal notranslate"><span class="pre">torch.device</span></code></a> as leftmost tensor in the list</p></li>
<li><p><strong>layout</strong> (<a class="reference internal" href="tensor_attributes.html#torch.layout" title="torch.layout"><code class="xref py py-class docutils literal notranslate"><span class="pre">torch.layout</span></code></a>, optional) – the desired layout of returned nested tensor.
Only strided and jagged layouts are supported. Default: if None, the strided layout.</p></li>
</ul>
</dd>
<dt class="field-odd">Return type</dt>
<dd class="field-odd"><p><a class="reference internal" href="tensors.html#torch.Tensor" title="torch.Tensor"><em>Tensor</em></a></p>
</dd>
</dl>
<p>Example:</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="gp">>>> </span><span class="n">a</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">arange</span><span class="p">(</span><span class="mi">3</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">requires_grad</span><span class="o">=</span><span class="kc">True</span><span class="p">)</span>
<span class="gp">>>> </span><span class="n">b</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">arange</span><span class="p">(</span><span class="mi">5</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">requires_grad</span><span class="o">=</span><span class="kc">True</span><span class="p">)</span>
<span class="gp">>>> </span><span class="n">nt</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">nested</span><span class="o">.</span><span class="n">as_nested_tensor</span><span class="p">([</span><span class="n">a</span><span class="p">,</span> <span class="n">b</span><span class="p">])</span>
<span class="gp">>>> </span><span class="n">nt</span><span class="o">.</span><span class="n">is_leaf</span>
<span class="go">False</span>
<span class="gp">>>> </span><span class="n">fake_grad</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">nested</span><span class="o">.</span><span class="n">nested_tensor</span><span class="p">([</span><span class="n">torch</span><span class="o">.</span><span class="n">ones_like</span><span class="p">(</span><span class="n">a</span><span class="p">),</span> <span class="n">torch</span><span class="o">.</span><span class="n">zeros_like</span><span class="p">(</span><span class="n">b</span><span class="p">)])</span>
<span class="gp">>>> </span><span class="n">nt</span><span class="o">.</span><span class="n">backward</span><span class="p">(</span><span class="n">fake_grad</span><span class="p">)</span>
<span class="gp">>>> </span><span class="n">a</span><span class="o">.</span><span class="n">grad</span>
<span class="go">tensor([1., 1., 1.])</span>
<span class="gp">>>> </span><span class="n">b</span><span class="o">.</span><span class="n">grad</span>
<span class="go">tensor([0., 0., 0., 0., 0.])</span>
<span class="gp">>>> </span><span class="n">c</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">randn</span><span class="p">(</span><span class="mi">3</span><span class="p">,</span> <span class="mi">5</span><span class="p">,</span> <span class="n">requires_grad</span><span class="o">=</span><span class="kc">True</span><span class="p">)</span>
<span class="gp">>>> </span><span class="n">nt2</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">nested</span><span class="o">.</span><span class="n">as_nested_tensor</span><span class="p">(</span><span class="n">c</span><span class="p">)</span>
</pre></div>
</div>
</dd></dl>
<dl class="py function">
<dt class="sig sig-object py" id="torch.nested.to_padded_tensor">
<span class="sig-prename descclassname"><span class="pre">torch.nested.</span></span><span class="sig-name descname"><span class="pre">to_padded_tensor</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">input</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">padding</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">output_size</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">out</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em><span class="sig-paren">)</span> <span class="sig-return"><span class="sig-return-icon">→</span> <span class="sig-return-typehint"><a class="reference internal" href="tensors.html#torch.Tensor" title="torch.Tensor"><span class="pre">Tensor</span></a></span></span><a class="headerlink" href="#torch.nested.to_padded_tensor" title="Permalink to this definition">¶</a></dt>
<dd><p>Returns a new (non-nested) Tensor by padding the <code class="xref py py-attr docutils literal notranslate"><span class="pre">input</span></code> nested tensor.
The leading entries will be filled with the nested data,
while the trailing entries will be padded.</p>
<div class="admonition warning">
<p class="admonition-title">Warning</p>
<p><a class="reference internal" href="#torch.nested.to_padded_tensor" title="torch.nested.to_padded_tensor"><code class="xref py py-func docutils literal notranslate"><span class="pre">to_padded_tensor()</span></code></a> always copies the underlying data,
since the nested and the non-nested tensors differ in memory layout.</p>
</div>
<dl class="field-list simple">
<dt class="field-odd">Parameters</dt>
<dd class="field-odd"><p><strong>padding</strong> (<a class="reference external" href="https://docs.python.org/3/library/functions.html#float" title="(in Python v3.13)"><em>float</em></a>) – The padding value for the trailing entries.</p>
</dd>
<dt class="field-even">Keyword Arguments</dt>
<dd class="field-even"><ul class="simple">
<li><p><strong>output_size</strong> (<em>Tuple</em><em>[</em><a class="reference external" href="https://docs.python.org/3/library/functions.html#int" title="(in Python v3.13)"><em>int</em></a><em>]</em>) – The size of the output tensor.
If given, it must be large enough to contain all nested data;
else, will infer by taking the max size of each nested sub-tensor along each dimension.</p></li>
<li><p><strong>out</strong> (<a class="reference internal" href="tensors.html#torch.Tensor" title="torch.Tensor"><em>Tensor</em></a><em>, </em><em>optional</em>) – the output tensor.</p></li>
</ul>
</dd>
</dl>
<p>Example:</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="gp">>>> </span><span class="n">nt</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">nested</span><span class="o">.</span><span class="n">nested_tensor</span><span class="p">([</span><span class="n">torch</span><span class="o">.</span><span class="n">randn</span><span class="p">((</span><span class="mi">2</span><span class="p">,</span> <span class="mi">5</span><span class="p">)),</span> <span class="n">torch</span><span class="o">.</span><span class="n">randn</span><span class="p">((</span><span class="mi">3</span><span class="p">,</span> <span class="mi">4</span><span class="p">))])</span>
<span class="go">nested_tensor([</span>
<span class="go"> tensor([[ 1.6862, -1.1282, 1.1031, 0.0464, -1.3276],</span>
<span class="go"> [-1.9967, -1.0054, 1.8972, 0.9174, -1.4995]]),</span>
<span class="go"> tensor([[-1.8546, -0.7194, -0.2918, -0.1846],</span>
<span class="go"> [ 0.2773, 0.8793, -0.5183, -0.6447],</span>
<span class="go"> [ 1.8009, 1.8468, -0.9832, -1.5272]])</span>
<span class="go">])</span>
<span class="gp">>>> </span><span class="n">pt_infer</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">nested</span><span class="o">.</span><span class="n">to_padded_tensor</span><span class="p">(</span><span class="n">nt</span><span class="p">,</span> <span class="mf">0.0</span><span class="p">)</span>
<span class="go">tensor([[[ 1.6862, -1.1282, 1.1031, 0.0464, -1.3276],</span>
<span class="go"> [-1.9967, -1.0054, 1.8972, 0.9174, -1.4995],</span>
<span class="go"> [ 0.0000, 0.0000, 0.0000, 0.0000, 0.0000]],</span>
<span class="go"> [[-1.8546, -0.7194, -0.2918, -0.1846, 0.0000],</span>
<span class="go"> [ 0.2773, 0.8793, -0.5183, -0.6447, 0.0000],</span>
<span class="go"> [ 1.8009, 1.8468, -0.9832, -1.5272, 0.0000]]])</span>
<span class="gp">>>> </span><span class="n">pt_large</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">nested</span><span class="o">.</span><span class="n">to_padded_tensor</span><span class="p">(</span><span class="n">nt</span><span class="p">,</span> <span class="mf">1.0</span><span class="p">,</span> <span class="p">(</span><span class="mi">2</span><span class="p">,</span> <span class="mi">4</span><span class="p">,</span> <span class="mi">6</span><span class="p">))</span>
<span class="go">tensor([[[ 1.6862, -1.1282, 1.1031, 0.0464, -1.3276, 1.0000],</span>
<span class="go"> [-1.9967, -1.0054, 1.8972, 0.9174, -1.4995, 1.0000],</span>
<span class="go"> [ 1.0000, 1.0000, 1.0000, 1.0000, 1.0000, 1.0000],</span>
<span class="go"> [ 1.0000, 1.0000, 1.0000, 1.0000, 1.0000, 1.0000]],</span>
<span class="go"> [[-1.8546, -0.7194, -0.2918, -0.1846, 1.0000, 1.0000],</span>
<span class="go"> [ 0.2773, 0.8793, -0.5183, -0.6447, 1.0000, 1.0000],</span>
<span class="go"> [ 1.8009, 1.8468, -0.9832, -1.5272, 1.0000, 1.0000],</span>
<span class="go"> [ 1.0000, 1.0000, 1.0000, 1.0000, 1.0000, 1.0000]]])</span>
<span class="gp">>>> </span><span class="n">pt_small</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">nested</span><span class="o">.</span><span class="n">to_padded_tensor</span><span class="p">(</span><span class="n">nt</span><span class="p">,</span> <span class="mf">2.0</span><span class="p">,</span> <span class="p">(</span><span class="mi">2</span><span class="p">,</span> <span class="mi">2</span><span class="p">,</span> <span class="mi">2</span><span class="p">))</span>
<span class="go">RuntimeError: Value in output_size is less than NestedTensor padded size. Truncation is not supported.</span>
</pre></div>
</div>
</dd></dl>
</div>
<div class="section" id="supported-operations">
<span id="id2"></span><h2>Supported operations<a class="headerlink" href="#supported-operations" title="Permalink to this heading">¶</a></h2>
<p>In this section, we summarize the operations that are currently supported on
NestedTensor and any constraints they have.</p>
<table class="colwidths-given docutils colwidths-auto align-default">
<colgroup>
<col style="width: 35%" />
<col style="width: 65%" />
</colgroup>
<thead>
<tr class="row-odd"><th class="head"><p>PyTorch operation</p></th>
<th class="head"><p>Constraints</p></th>
</tr>
</thead>
<tbody>
<tr class="row-even"><td><p><a class="reference internal" href="generated/torch.matmul.html#torch.matmul" title="torch.matmul"><code class="xref py py-func docutils literal notranslate"><span class="pre">torch.matmul()</span></code></a></p></td>
<td><p>Supports matrix multiplication between two (>= 3d) nested tensors where
the last two dimensions are matrix dimensions and the leading (batch) dimensions have the same size
(i.e. no broadcasting support for batch dimensions yet).</p></td>
</tr>
<tr class="row-odd"><td><p><a class="reference internal" href="generated/torch.bmm.html#torch.bmm" title="torch.bmm"><code class="xref py py-func docutils literal notranslate"><span class="pre">torch.bmm()</span></code></a></p></td>
<td><p>Supports batch matrix multiplication of two 3-d nested tensors.</p></td>
</tr>
<tr class="row-even"><td><p><a class="reference internal" href="generated/torch.nn.Linear.html#torch.nn.Linear" title="torch.nn.Linear"><code class="xref py py-func docutils literal notranslate"><span class="pre">torch.nn.Linear()</span></code></a></p></td>
<td><p>Supports 3-d nested input and a dense 2-d weight matrix.</p></td>
</tr>
<tr class="row-odd"><td><p><a class="reference internal" href="generated/torch.nn.functional.softmax.html#torch.nn.functional.softmax" title="torch.nn.functional.softmax"><code class="xref py py-func docutils literal notranslate"><span class="pre">torch.nn.functional.softmax()</span></code></a></p></td>
<td><p>Supports softmax along all dims except dim=0.</p></td>
</tr>
<tr class="row-even"><td><p><a class="reference internal" href="generated/torch.nn.Dropout.html#torch.nn.Dropout" title="torch.nn.Dropout"><code class="xref py py-func docutils literal notranslate"><span class="pre">torch.nn.Dropout()</span></code></a></p></td>
<td><p>Behavior is the same as on regular tensors.</p></td>
</tr>
<tr class="row-odd"><td><p><a class="reference internal" href="generated/torch.Tensor.masked_fill.html#torch.Tensor.masked_fill" title="torch.Tensor.masked_fill"><code class="xref py py-func docutils literal notranslate"><span class="pre">torch.Tensor.masked_fill()</span></code></a></p></td>
<td><p>Behavior is the same as on regular tensors.</p></td>
</tr>
<tr class="row-even"><td><p><code class="xref py py-func docutils literal notranslate"><span class="pre">torch.relu()</span></code></p></td>
<td><p>Behavior is the same as on regular tensors.</p></td>
</tr>
<tr class="row-odd"><td><p><code class="xref py py-func docutils literal notranslate"><span class="pre">torch.gelu()</span></code></p></td>
<td><p>Behavior is the same as on regular tensors.</p></td>
</tr>
<tr class="row-even"><td><p><code class="xref py py-func docutils literal notranslate"><span class="pre">torch.silu()</span></code></p></td>
<td><p>Behavior is the same as on regular tensors.</p></td>
</tr>
<tr class="row-odd"><td><p><a class="reference internal" href="generated/torch.abs.html#torch.abs" title="torch.abs"><code class="xref py py-func docutils literal notranslate"><span class="pre">torch.abs()</span></code></a></p></td>
<td><p>Behavior is the same as on regular tensors.</p></td>
</tr>
<tr class="row-even"><td><p><a class="reference internal" href="generated/torch.sgn.html#torch.sgn" title="torch.sgn"><code class="xref py py-func docutils literal notranslate"><span class="pre">torch.sgn()</span></code></a></p></td>
<td><p>Behavior is the same as on regular tensors.</p></td>
</tr>
<tr class="row-odd"><td><p><a class="reference internal" href="generated/torch.logical_not.html#torch.logical_not" title="torch.logical_not"><code class="xref py py-func docutils literal notranslate"><span class="pre">torch.logical_not()</span></code></a></p></td>
<td><p>Behavior is the same as on regular tensors.</p></td>
</tr>
<tr class="row-even"><td><p><a class="reference internal" href="generated/torch.neg.html#torch.neg" title="torch.neg"><code class="xref py py-func docutils literal notranslate"><span class="pre">torch.neg()</span></code></a></p></td>
<td><p>Behavior is the same as on regular tensors.</p></td>
</tr>
<tr class="row-odd"><td><p><a class="reference internal" href="generated/torch.sub.html#torch.sub" title="torch.sub"><code class="xref py py-func docutils literal notranslate"><span class="pre">torch.sub()</span></code></a></p></td>
<td><p>Supports elementwise subtraction of two nested tensors.</p></td>
</tr>
<tr class="row-even"><td><p><a class="reference internal" href="generated/torch.add.html#torch.add" title="torch.add"><code class="xref py py-func docutils literal notranslate"><span class="pre">torch.add()</span></code></a></p></td>
<td><p>Supports elementwise addition of two nested tensors. Supports addition of a scalar to a nested tensor.</p></td>
</tr>
<tr class="row-odd"><td><p><a class="reference internal" href="generated/torch.mul.html#torch.mul" title="torch.mul"><code class="xref py py-func docutils literal notranslate"><span class="pre">torch.mul()</span></code></a></p></td>
<td><p>Supports elementwise multiplication of two nested tensors. Supports multiplication of a nested tensor by a scalar.</p></td>
</tr>
<tr class="row-even"><td><p><a class="reference internal" href="generated/torch.select.html#torch.select" title="torch.select"><code class="xref py py-func docutils literal notranslate"><span class="pre">torch.select()</span></code></a></p></td>
<td><p>Supports selecting along all dimensions.</p></td>
</tr>
<tr class="row-odd"><td><p><a class="reference internal" href="generated/torch.clone.html#torch.clone" title="torch.clone"><code class="xref py py-func docutils literal notranslate"><span class="pre">torch.clone()</span></code></a></p></td>
<td><p>Behavior is the same as on regular tensors.</p></td>
</tr>
<tr class="row-even"><td><p><code class="xref py py-func docutils literal notranslate"><span class="pre">torch.detach()</span></code></p></td>
<td><p>Behavior is the same as on regular tensors.</p></td>
</tr>
<tr class="row-odd"><td><p><a class="reference internal" href="generated/torch.unbind.html#torch.unbind" title="torch.unbind"><code class="xref py py-func docutils literal notranslate"><span class="pre">torch.unbind()</span></code></a></p></td>
<td><p>Supports unbinding along <code class="docutils literal notranslate"><span class="pre">dim=0</span></code> only.</p></td>
</tr>
<tr class="row-even"><td><p><a class="reference internal" href="generated/torch.reshape.html#torch.reshape" title="torch.reshape"><code class="xref py py-func docutils literal notranslate"><span class="pre">torch.reshape()</span></code></a></p></td>
<td><p>Supports reshaping with size of <code class="docutils literal notranslate"><span class="pre">dim=0</span></code> preserved (i.e. number of tensors nested cannot be changed).
Unlike regular tensors, a size of <code class="docutils literal notranslate"><span class="pre">-1</span></code> here means that the existing size is inherited.
In particular, the only valid size for a irregular dimension is <code class="docutils literal notranslate"><span class="pre">-1</span></code>.
Size inference is not implemented yet and hence for new dimensions the size cannot be <code class="docutils literal notranslate"><span class="pre">-1</span></code>.</p></td>
</tr>
<tr class="row-odd"><td><p><a class="reference internal" href="generated/torch.Tensor.reshape_as.html#torch.Tensor.reshape_as" title="torch.Tensor.reshape_as"><code class="xref py py-func docutils literal notranslate"><span class="pre">torch.Tensor.reshape_as()</span></code></a></p></td>
<td><p>Similar constraint as for <code class="docutils literal notranslate"><span class="pre">reshape</span></code>.</p></td>
</tr>
<tr class="row-even"><td><p><a class="reference internal" href="generated/torch.transpose.html#torch.transpose" title="torch.transpose"><code class="xref py py-func docutils literal notranslate"><span class="pre">torch.transpose()</span></code></a></p></td>
<td><p>Supports transposing of all dims except <code class="docutils literal notranslate"><span class="pre">dim=0</span></code>.</p></td>
</tr>
<tr class="row-odd"><td><p><a class="reference internal" href="generated/torch.Tensor.view.html#torch.Tensor.view" title="torch.Tensor.view"><code class="xref py py-func docutils literal notranslate"><span class="pre">torch.Tensor.view()</span></code></a></p></td>
<td><p>Rules for the new shape are similar to that of <code class="docutils literal notranslate"><span class="pre">reshape</span></code>.</p></td>
</tr>
<tr class="row-even"><td><p><a class="reference internal" href="generated/torch.empty_like.html#torch.empty_like" title="torch.empty_like"><code class="xref py py-func docutils literal notranslate"><span class="pre">torch.empty_like()</span></code></a></p></td>
<td><p>Behavior is analogous to that of regular tensors; returns a new empty nested tensor (i.e. with uninitialized values) matching the nested structure of the input.</p></td>
</tr>
<tr class="row-odd"><td><p><a class="reference internal" href="generated/torch.randn_like.html#torch.randn_like" title="torch.randn_like"><code class="xref py py-func docutils literal notranslate"><span class="pre">torch.randn_like()</span></code></a></p></td>
<td><p>Behavior is analogous to that of regular tensors; returns a new nested tensor with values randomly initialized according to a standard normal distribution matching the nested structure of the input.</p></td>
</tr>
<tr class="row-even"><td><p><a class="reference internal" href="generated/torch.zeros_like.html#torch.zeros_like" title="torch.zeros_like"><code class="xref py py-func docutils literal notranslate"><span class="pre">torch.zeros_like()</span></code></a></p></td>
<td><p>Behavior is analogous to that of regular tensors; returns a new nested tensor with all zero values matching the nested structure of the input.</p></td>
</tr>
<tr class="row-odd"><td><p><a class="reference internal" href="generated/torch.nn.LayerNorm.html#torch.nn.LayerNorm" title="torch.nn.LayerNorm"><code class="xref py py-func docutils literal notranslate"><span class="pre">torch.nn.LayerNorm()</span></code></a></p></td>
<td><p>The <code class="docutils literal notranslate"><span class="pre">normalized_shape</span></code> argument is restricted to not extend into the irregular dimensions of the NestedTensor.</p></td>
</tr>
</tbody>
</table>
</div>
</div>
</article>
</div>
<footer>
<div class="rst-footer-buttons" role="navigation" aria-label="footer navigation">
<a href="size.html" class="btn btn-neutral float-right" title="torch.Size" accesskey="n" rel="next">Next <img src="_static/images/chevron-right-orange.svg" class="next-page"></a>
<a href="masked.html" class="btn btn-neutral" title="torch.masked" accesskey="p" rel="prev"><img src="_static/images/chevron-right-orange.svg" class="previous-page"> Previous</a>
</div>
<hr>
<div role="contentinfo">
<p>
© Copyright 2023, PyTorch Contributors.
</p>
</div>
<div>
Built with <a href="http://sphinx-doc.org/">Sphinx</a> using a <a href="https://github.com/rtfd/sphinx_rtd_theme">theme</a> provided by <a href="https://readthedocs.org">Read the Docs</a>.
</div>
</footer>
</div>
<script>
var match = window.location.href.match(/\/_[a-zA-Z0-9_]*.html|_dynamo/gi);
var url = window.location.href.lastIndexOf(match[match.length-1]);
if (url)
{
var div = '<div class="admonition note"><p class="admonition-title">Note</p><p><i class="fa fa-exclamation-circle" aria-hidden="true"> </i> This page describes an internal API which is not intended to be used outside of the PyTorch codebase and can be modified or removed without notice.</p></div>'
document.getElementById("pytorch-article").insertAdjacentHTML('afterBegin', div)
}
</script>
</div>
<div class="pytorch-content-right" id="pytorch-content-right">
<div class="pytorch-right-menu" id="pytorch-right-menu">
<div class="pytorch-side-scroll" id="pytorch-side-scroll-right">
<ul>
<li><a class="reference internal" href="#">torch.nested</a><ul>
<li><a class="reference internal" href="#introduction">Introduction</a></li>
<li><a class="reference internal" href="#construction">Construction</a></li>
<li><a class="reference internal" href="#size">size</a></li>
<li><a class="reference internal" href="#unbind">unbind</a></li>
<li><a class="reference internal" href="#nested-tensor-constructor-and-conversion-functions">Nested tensor constructor and conversion functions</a><ul>
<li><a class="reference internal" href="#torch.nested.nested_tensor"><code class="docutils literal notranslate"><span class="pre">nested_tensor()</span></code></a></li>
<li><a class="reference internal" href="#torch.nested.as_nested_tensor"><code class="docutils literal notranslate"><span class="pre">as_nested_tensor()</span></code></a></li>
<li><a class="reference internal" href="#torch.nested.to_padded_tensor"><code class="docutils literal notranslate"><span class="pre">to_padded_tensor()</span></code></a></li>
</ul>
</li>
<li><a class="reference internal" href="#supported-operations">Supported operations</a></li>
</ul>
</li>
</ul>
</div>
</div>
</div>
</section>
</div>