-
Notifications
You must be signed in to change notification settings - Fork 362
/
Copy pathgetting_started_with_cpp_api.html
1143 lines (951 loc) · 87.5 KB
/
getting_started_with_cpp_api.html
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
<!DOCTYPE html>
<!--[if IE 8]><html class="no-js lt-ie9" lang="en" > <![endif]-->
<!--[if gt IE 8]><!--> <html class="no-js" lang="en" > <!--<![endif]-->
<head>
<meta charset="utf-8">
<meta name="generator" content="Docutils 0.17.1: http://docutils.sourceforge.net/" />
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>Using Torch-TensorRT in C++ — Torch-TensorRT v2.7.0.dev0+f34e667 documentation</title>
<link rel="stylesheet" href="../_static/css/theme.css" type="text/css" />
<!-- <link rel="stylesheet" href="../_static/pygments.css" type="text/css" /> -->
<link rel="stylesheet" href="../_static/pygments.css" type="text/css" />
<link rel="stylesheet" href="../_static/css/theme.css" type="text/css" />
<link rel="stylesheet" href="../_static/sg_gallery.css" type="text/css" />
<link rel="stylesheet" href="../_static/sg_gallery-binder.css" type="text/css" />
<link rel="stylesheet" href="../_static/sg_gallery-dataframe.css" type="text/css" />
<link rel="stylesheet" href="../_static/sg_gallery-rendered-html.css" type="text/css" />
<link rel="stylesheet" href="../_static/collapsible-lists/css/tree_view.css" type="text/css" />
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/katex@0.10.0-beta/dist/katex.min.css" type="text/css" />
<link rel="stylesheet" href="../_static/css/custom.css" type="text/css" />
<link rel="index" title="Index" href="../genindex.html" />
<link rel="search" title="Search" href="../search.html" />
<link rel="next" title="Post Training Quantization (PTQ)" href="ptq.html" />
<link rel="prev" title="Using Torch-TensorRT in Python" href="getting_started_with_python_api.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','');</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/katex@0.10.0/dist/fonts/KaTeX_Math-Italic.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="preload" href="https://cdn.jsdelivr.net/npm/katex@0.10.0/dist/fonts/KaTeX_Main-Regular.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="preload" href="https://cdn.jsdelivr.net/npm/katex@0.10.0/dist/fonts/KaTeX_Main-Bold.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="preload" href="https://cdn.jsdelivr.net/npm/katex@0.10.0/dist/fonts/KaTeX_Size1-Regular.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="preload" href="https://cdn.jsdelivr.net/npm/katex@0.10.0/dist/fonts/KaTeX_Size4-Regular.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="preload" href="https://cdn.jsdelivr.net/npm/katex@0.10.0/dist/fonts/KaTeX_Size2-Regular.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="preload" href="https://cdn.jsdelivr.net/npm/katex@0.10.0/dist/fonts/KaTeX_Size3-Regular.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="preload" href="https://cdn.jsdelivr.net/npm/katex@0.10.0/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-2024">
<span class="dropdown-title">Contributor Awards - 2024</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>
<a class="nav-dropdown-item" href="https://pytorch.org/executorch/stable/index.html">
<span class="dropdown-title">ExecuTorch Docs</span>
</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>
<a class="nav-dropdown-item" href="https://pytorch.org/newsletter">
<span class="dropdown-title">Newsletter</span>
<p>Stay up-to-date with the latest updates</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>
</a>
<a class="nav-dropdown-item" href="https://pytorch.org/credits">
<span class="dropdown-title">Cloud Credit Program</span>
</a>
<a class="nav-dropdown-item" href="https://pytorch.org/tac">
<span class="dropdown-title">Technical Advisory Council</span>
</a>
<a class="nav-dropdown-item" href="https://pytorch.org/staff">
<span class="dropdown-title">Staff</span>
</a>
<a class="nav-dropdown-item" href="https://pytorch.org/contact-us">
<span class="dropdown-title">Contact Us</span>
</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">
v2.7.0.dev0+f34e667
</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">Getting Started</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../getting_started/installation.html">Installation</a></li>
<li class="toctree-l1"><a class="reference internal" href="../getting_started/jetpack.html">Overview</a></li>
<li class="toctree-l1"><a class="reference internal" href="../getting_started/quick_start.html">Quick Start</a></li>
</ul>
<p class="caption" role="heading"><span class="caption-text">User Guide</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../user_guide/torch_tensorrt_explained.html">Torch-TensorRT Explained</a></li>
<li class="toctree-l1"><a class="reference internal" href="../user_guide/dynamic_shapes.html">Dynamic shapes with Torch-TensorRT</a></li>
<li class="toctree-l1"><a class="reference internal" href="../user_guide/saving_models.html">Saving models compiled with Torch-TensorRT</a></li>
<li class="toctree-l1"><a class="reference internal" href="../user_guide/runtime.html">Deploying Torch-TensorRT Programs</a></li>
<li class="toctree-l1"><a class="reference internal" href="../user_guide/using_dla.html">DLA</a></li>
<li class="toctree-l1"><a class="reference internal" href="../user_guide/mixed_precision.html">Compile Mixed Precision models with Torch-TensorRT</a></li>
</ul>
<p class="caption" role="heading"><span class="caption-text">Tutorials</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../tutorials/_rendered_examples/dynamo/torch_compile_advanced_usage.html">Torch Compile Advanced Usage</a></li>
<li class="toctree-l1"><a class="reference internal" href="../tutorials/_rendered_examples/dynamo/vgg16_ptq.html">Deploy Quantized Models using Torch-TensorRT</a></li>
<li class="toctree-l1"><a class="reference internal" href="../tutorials/_rendered_examples/dynamo/engine_caching_example.html">Engine Caching</a></li>
<li class="toctree-l1"><a class="reference internal" href="../tutorials/_rendered_examples/dynamo/engine_caching_bert_example.html">Engine Caching (BERT)</a></li>
<li class="toctree-l1"><a class="reference internal" href="../tutorials/_rendered_examples/dynamo/refit_engine_example.html">Refitting Torch-TensorRT Programs with New Weights</a></li>
<li class="toctree-l1"><a class="reference internal" href="../tutorials/serving_torch_tensorrt_with_triton.html">Serving a Torch-TensorRT model with Triton</a></li>
<li class="toctree-l1"><a class="reference internal" href="../tutorials/_rendered_examples/dynamo/torch_export_cudagraphs.html">Torch Export with Cudagraphs</a></li>
<li class="toctree-l1"><a class="reference internal" href="../tutorials/_rendered_examples/dynamo/converter_overloading.html">Overloading Torch-TensorRT Converters with Custom Converters</a></li>
<li class="toctree-l1"><a class="reference internal" href="../tutorials/_rendered_examples/dynamo/custom_kernel_plugins.html">Using Custom Kernels within TensorRT Engines with Torch-TensorRT</a></li>
<li class="toctree-l1"><a class="reference internal" href="../tutorials/_rendered_examples/dynamo/auto_generate_converters.html">Automatically Generate a Converter for a Custom Kernel</a></li>
<li class="toctree-l1"><a class="reference internal" href="../tutorials/_rendered_examples/dynamo/auto_generate_plugins.html">Automatically Generate a Plugin for a Custom Kernel</a></li>
<li class="toctree-l1"><a class="reference internal" href="../tutorials/_rendered_examples/dynamo/mutable_torchtrt_module_example.html">Mutable Torch TensorRT Module</a></li>
<li class="toctree-l1"><a class="reference internal" href="../tutorials/_rendered_examples/dynamo/weight_streaming_example.html">Weight Streaming</a></li>
<li class="toctree-l1"><a class="reference internal" href="../tutorials/_rendered_examples/dynamo/pre_allocated_output_example.html">Pre-allocated output buffer</a></li>
</ul>
<p class="caption" role="heading"><span class="caption-text">Dynamo Frontend</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../dynamo/torch_compile.html">TensorRT Backend for <code class="docutils literal notranslate"><span class="pre">torch.compile</span></code></a></li>
<li class="toctree-l1"><a class="reference internal" href="../dynamo/dynamo_export.html">Compiling Exported Programs with Torch-TensorRT</a></li>
</ul>
<p class="caption" role="heading"><span class="caption-text">TorchScript Frontend</span></p>
<ul class="current">
<li class="toctree-l1"><a class="reference internal" href="creating_torchscript_module_in_python.html">Creating a TorchScript Module</a></li>
<li class="toctree-l1"><a class="reference internal" href="creating_torchscript_module_in_python.html#working-with-torchscript-in-python">Working with TorchScript in Python</a></li>
<li class="toctree-l1"><a class="reference internal" href="creating_torchscript_module_in_python.html#saving-torchscript-module-to-disk">Saving TorchScript Module to Disk</a></li>
<li class="toctree-l1"><a class="reference internal" href="getting_started_with_python_api.html">Using Torch-TensorRT in Python</a></li>
<li class="toctree-l1 current"><a class="current reference internal" href="#">Using Torch-TensorRT in C++</a></li>
<li class="toctree-l1"><a class="reference internal" href="ptq.html">Post Training Quantization (PTQ)</a></li>
</ul>
<p class="caption" role="heading"><span class="caption-text">FX Frontend</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../fx/getting_started_with_fx_path.html">Torch-TensorRT (FX Frontend) User Guide</a></li>
</ul>
<p class="caption" role="heading"><span class="caption-text">Model Zoo</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../tutorials/_rendered_examples/dynamo/torch_compile_resnet_example.html">Compiling ResNet with dynamic shapes using the <cite>torch.compile</cite> backend</a></li>
<li class="toctree-l1"><a class="reference internal" href="../tutorials/_rendered_examples/dynamo/torch_compile_transformers_example.html">Compiling BERT using the <cite>torch.compile</cite> backend</a></li>
<li class="toctree-l1"><a class="reference internal" href="../tutorials/_rendered_examples/dynamo/torch_compile_stable_diffusion.html">Compiling Stable Diffusion model using the <cite>torch.compile</cite> backend</a></li>
<li class="toctree-l1"><a class="reference internal" href="../tutorials/_rendered_examples/dynamo/torch_compile_gpt2.html">Compiling GPT2 using the Torch-TensorRT <code class="docutils literal notranslate"><span class="pre">torch.compile</span></code> frontend</a></li>
<li class="toctree-l1"><a class="reference internal" href="../tutorials/_rendered_examples/dynamo/torch_export_gpt2.html">Compiling GPT2 using the dynamo backend</a></li>
<li class="toctree-l1"><a class="reference internal" href="../tutorials/_rendered_examples/dynamo/torch_export_llama2.html">Compiling Llama2 using the dynamo backend</a></li>
<li class="toctree-l1"><a class="reference internal" href="../tutorials/_rendered_examples/dynamo/torch_export_sam2.html">Compiling SAM2 using the dynamo backend</a></li>
<li class="toctree-l1"><a class="reference internal" href="../tutorials/_rendered_examples/dynamo/torch_export_flux_dev.html">Compiling FLUX.1-dev model using the Torch-TensorRT dynamo backend</a></li>
<li class="toctree-l1"><a class="reference internal" href="../tutorials/notebooks.html">Legacy notebooks</a></li>
</ul>
<p class="caption" role="heading"><span class="caption-text">Python API Documentation</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../py_api/torch_tensorrt.html">torch_tensorrt</a></li>
<li class="toctree-l1"><a class="reference internal" href="../py_api/dynamo.html">torch_tensorrt.dynamo</a></li>
<li class="toctree-l1"><a class="reference internal" href="../py_api/logging.html">torch_tensorrt.logging</a></li>
<li class="toctree-l1"><a class="reference internal" href="../py_api/fx.html">torch_tensorrt.fx</a></li>
<li class="toctree-l1"><a class="reference internal" href="../py_api/ts.html">torch_tensorrt.ts</a></li>
<li class="toctree-l1"><a class="reference internal" href="../py_api/ptq.html">torch_tensorrt.ts.ptq</a></li>
</ul>
<p class="caption" role="heading"><span class="caption-text">C++ API Documentation</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../_cpp_api/torch_tensort_cpp.html">Torch-TensorRT C++ API</a></li>
<li class="toctree-l1"><a class="reference internal" href="../_cpp_api/namespace_torch_tensorrt.html">Namespace torch_tensorrt</a></li>
<li class="toctree-l1"><a class="reference internal" href="../_cpp_api/namespace_torch_tensorrt__logging.html">Namespace torch_tensorrt::logging</a></li>
<li class="toctree-l1"><a class="reference internal" href="../_cpp_api/namespace_torch_tensorrt__torchscript.html">Namespace torch_tensorrt::torchscript</a></li>
<li class="toctree-l1"><a class="reference internal" href="../_cpp_api/namespace_torch_tensorrt__ptq.html">Namespace torch_tensorrt::ptq</a></li>
</ul>
<p class="caption" role="heading"><span class="caption-text">CLI Documentation</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../cli/torchtrtc.html">torchtrtc</a></li>
</ul>
<p class="caption" role="heading"><span class="caption-text">Contributor Documentation</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../contributors/system_overview.html">System Overview</a></li>
<li class="toctree-l1"><a class="reference internal" href="../contributors/dynamo_converters.html">Writing Dynamo Converters</a></li>
<li class="toctree-l1"><a class="reference internal" href="../contributors/writing_dynamo_aten_lowering_passes.html">Writing Dynamo ATen Lowering Passes</a></li>
<li class="toctree-l1"><a class="reference internal" href="../contributors/ts_converters.html">Writing TorchScript Converters</a></li>
<li class="toctree-l1"><a class="reference internal" href="../contributors/useful_links.html">Useful Links for Torch-TensorRT Development</a></li>
</ul>
<p class="caption" role="heading"><span class="caption-text">Indices</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../indices/supported_ops.html">Operators Supported</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>Using Torch-TensorRT in C++</li>
<li class="pytorch-breadcrumbs-aside">
<a href="../_sources/ts/getting_started_with_cpp_api.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="
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">
<section id="using-torch-tensorrt-in-c">
<span id="getting-started-cpp"></span><h1>Using Torch-TensorRT in C++<a class="headerlink" href="#using-torch-tensorrt-in-c" title="Permalink to this heading">¶</a></h1>
<p>If you haven’t already, acquire a tarball of the library by following the instructions in <a class="reference internal" href="../getting_started/installation.html#installation"><span class="std std-ref">Installation</span></a></p>
<section id="id1">
<h2>Using Torch-TensorRT in C++<a class="headerlink" href="#id1" title="Permalink to this heading">¶</a></h2>
<p>Torch-TensorRT C++ API accepts TorchScript modules (generated either from <code class="docutils literal notranslate"><span class="pre">torch.jit.script</span></code> or <code class="docutils literal notranslate"><span class="pre">torch.jit.trace</span></code>) as an input and returns
a Torchscript module (optimized using TensorRT), Dynamo compilation workflows will not be supported in the C++ API however, execution of
torch.jit.trace’d compiled FX GraphModules is supported for FX and Dyanmo workflows.</p>
<p>Please refer to <a class="reference external" href="https://nvidia.github.io/Torch-TensorRT/tutorials/creating_torchscript_module_in_python.html">Creating TorchScript modules in Python</a> section to generate torchscript graphs.</p>
<section id="torch-tensorrt-quickstart-compiling-torchscript-modules-with-torchtrtc">
<span id="torch-tensorrt-quickstart"></span><h3>[Torch-TensorRT Quickstart] Compiling TorchScript Modules with <code class="docutils literal notranslate"><span class="pre">torchtrtc</span></code><a class="headerlink" href="#torch-tensorrt-quickstart-compiling-torchscript-modules-with-torchtrtc" title="Permalink to this heading">¶</a></h3>
<p>An easy way to get started with Torch-TensorRT and to check if your model can be supported without extra work is to run it through
<code class="docutils literal notranslate"><span class="pre">torchtrtc</span></code>, which supports almost all features of the compiler from the command line including post training quantization
(given a previously created calibration cache). For example we can compile our lenet model by setting our preferred operating
precision and input size. This new TorchScript file can be loaded into Python (note: you need to <code class="docutils literal notranslate"><span class="pre">import</span> <span class="pre">torch_tensorrt</span></code> before loading
these compiled modules because the compiler extends the PyTorch the deserializer and runtime to execute compiled modules).</p>
<div class="highlight-shell notranslate"><div class="highlight"><pre><span></span>❯<span class="w"> </span>torchtrtc<span class="w"> </span>-p<span class="w"> </span>f16<span class="w"> </span>lenet_scripted.ts<span class="w"> </span>trt_lenet_scripted.ts<span class="w"> </span><span class="s2">"(1,1,32,32)"</span>
❯<span class="w"> </span>python3
Python<span class="w"> </span><span class="m">3</span>.6.9<span class="w"> </span><span class="o">(</span>default,<span class="w"> </span>Apr<span class="w"> </span><span class="m">18</span><span class="w"> </span><span class="m">2020</span>,<span class="w"> </span><span class="m">01</span>:56:04<span class="o">)</span>
<span class="o">[</span>GCC<span class="w"> </span><span class="m">8</span>.4.0<span class="o">]</span><span class="w"> </span>on<span class="w"> </span>linux
Type<span class="w"> </span><span class="s2">"help"</span>,<span class="w"> </span><span class="s2">"copyright"</span>,<span class="w"> </span><span class="s2">"credits"</span><span class="w"> </span>or<span class="w"> </span><span class="s2">"license"</span><span class="w"> </span><span class="k">for</span><span class="w"> </span>more<span class="w"> </span>information.
>>><span class="w"> </span>import<span class="w"> </span>torch
>>><span class="w"> </span>import<span class="w"> </span>torch_tensorrt
>>><span class="w"> </span><span class="nv">ts_model</span><span class="w"> </span><span class="o">=</span><span class="w"> </span>torch.jit.load<span class="o">(</span>“trt_lenet_scripted.ts”<span class="o">)</span>
>>><span class="w"> </span>ts_model<span class="o">(</span>torch.randn<span class="o">((</span><span class="m">1</span>,1,32,32<span class="o">))</span>.to<span class="o">(</span>“cuda”<span class="o">)</span>.half<span class="o">())</span>
</pre></div>
</div>
<p>You can learn more about <code class="docutils literal notranslate"><span class="pre">torchtrtc</span></code> usage here: <a class="reference internal" href="../cli/torchtrtc.html#torchtrtc"><span class="std std-ref">torchtrtc</span></a></p>
</section>
<section id="working-with-torchscript-in-c">
<span id="ts-in-cc"></span><h3>Working with TorchScript in C++<a class="headerlink" href="#working-with-torchscript-in-c" title="Permalink to this heading">¶</a></h3>
<p>If we are developing an application to deploy with C++, we can save either our traced or scripted module using <code class="docutils literal notranslate"><span class="pre">torch.jit.save</span></code>
which will serialize the TorchScript code, weights and other information into a package. This is also where our dependency on Python ends.</p>
<div class="highlight-python notranslate"><div class="highlight"><pre><span></span><span class="n">torch_script_module</span><span class="o">.</span><span class="n">save</span><span class="p">(</span><span class="s2">"lenet.jit.pt"</span><span class="p">)</span>
</pre></div>
</div>
<p>From here we can now load our TorchScript module in C++</p>
<div class="highlight-c++ notranslate"><div class="highlight"><pre><span></span><span class="cp">#include</span><span class="w"> </span><span class="cpf"><torch/script.h></span><span class="c1"> // One-stop header.</span>
<span class="cp">#include</span><span class="w"> </span><span class="cpf"><iostream></span>
<span class="cp">#include</span><span class="w"> </span><span class="cpf"><memory></span>
<span class="kt">int</span><span class="w"> </span><span class="nf">main</span><span class="p">(</span><span class="kt">int</span><span class="w"> </span><span class="n">argc</span><span class="p">,</span><span class="w"> </span><span class="k">const</span><span class="w"> </span><span class="kt">char</span><span class="o">*</span><span class="w"> </span><span class="n">argv</span><span class="p">[])</span><span class="w"> </span><span class="p">{</span>
<span class="w"> </span><span class="n">torch</span><span class="o">::</span><span class="n">jit</span><span class="o">::</span><span class="n">Module</span><span class="w"> </span><span class="k">module</span><span class="p">;</span>
<span class="w"> </span><span class="k">try</span><span class="w"> </span><span class="p">{</span>
<span class="w"> </span><span class="c1">// Deserialize the ScriptModule from a file using torch::jit::load().</span>
<span class="w"> </span><span class="k">module</span><span class="w"> </span><span class="o">=</span><span class="w"> </span><span class="n">torch</span><span class="o">::</span><span class="n">jit</span><span class="o">::</span><span class="n">load</span><span class="p">(</span><span class="s">"<PATH TO SAVED TS MOD>"</span><span class="p">);</span>
<span class="w"> </span><span class="p">}</span>
<span class="w"> </span><span class="k">catch</span><span class="w"> </span><span class="p">(</span><span class="k">const</span><span class="w"> </span><span class="n">c10</span><span class="o">::</span><span class="n">Error</span><span class="o">&</span><span class="w"> </span><span class="n">e</span><span class="p">)</span><span class="w"> </span><span class="p">{</span>
<span class="w"> </span><span class="n">std</span><span class="o">::</span><span class="n">cerr</span><span class="w"> </span><span class="o"><<</span><span class="w"> </span><span class="s">"error loading the model</span><span class="se">\n</span><span class="s">"</span><span class="p">;</span>
<span class="w"> </span><span class="k">return</span><span class="w"> </span><span class="mi">-1</span><span class="p">;</span>
<span class="w"> </span><span class="p">}</span>
<span class="w"> </span><span class="n">std</span><span class="o">::</span><span class="n">cout</span><span class="w"> </span><span class="o"><<</span><span class="w"> </span><span class="s">"ok</span><span class="se">\n</span><span class="s">"</span><span class="p">;</span>
</pre></div>
</div>
<p>You can do full training and inference in C++ with PyTorch / LibTorch if you would like, you can even define your modules in C++ and
have access to the same powerful tensor library that backs PyTorch. (For more information: <a class="reference external" href="https://pytorch.org/cppdocs/">https://pytorch.org/cppdocs/</a>).
For instance we can do inference with our LeNet module like this:</p>
<div class="highlight-c++ notranslate"><div class="highlight"><pre><span></span><span class="n">mod</span><span class="p">.</span><span class="n">eval</span><span class="p">();</span>
<span class="n">torch</span><span class="o">::</span><span class="n">Tensor</span><span class="w"> </span><span class="n">in</span><span class="w"> </span><span class="o">=</span><span class="w"> </span><span class="n">torch</span><span class="o">::</span><span class="n">randn</span><span class="p">({</span><span class="mi">1</span><span class="p">,</span><span class="w"> </span><span class="mi">1</span><span class="p">,</span><span class="w"> </span><span class="mi">32</span><span class="p">,</span><span class="w"> </span><span class="mi">32</span><span class="p">});</span>
<span class="k">auto</span><span class="w"> </span><span class="n">out</span><span class="w"> </span><span class="o">=</span><span class="w"> </span><span class="n">mod</span><span class="p">.</span><span class="n">forward</span><span class="p">(</span><span class="n">in</span><span class="p">);</span>
</pre></div>
</div>
<p>and to run on the GPU:</p>
<div class="highlight-c++ notranslate"><div class="highlight"><pre><span></span><span class="n">mod</span><span class="p">.</span><span class="n">eval</span><span class="p">();</span>
<span class="n">mod</span><span class="p">.</span><span class="n">to</span><span class="p">(</span><span class="n">torch</span><span class="o">::</span><span class="n">kCUDA</span><span class="p">);</span>
<span class="n">torch</span><span class="o">::</span><span class="n">Tensor</span><span class="w"> </span><span class="n">in</span><span class="w"> </span><span class="o">=</span><span class="w"> </span><span class="n">torch</span><span class="o">::</span><span class="n">randn</span><span class="p">({</span><span class="mi">1</span><span class="p">,</span><span class="w"> </span><span class="mi">1</span><span class="p">,</span><span class="w"> </span><span class="mi">32</span><span class="p">,</span><span class="w"> </span><span class="mi">32</span><span class="p">},</span><span class="w"> </span><span class="n">torch</span><span class="o">::</span><span class="n">kCUDA</span><span class="p">);</span>
<span class="k">auto</span><span class="w"> </span><span class="n">out</span><span class="w"> </span><span class="o">=</span><span class="w"> </span><span class="n">mod</span><span class="p">.</span><span class="n">forward</span><span class="p">(</span><span class="n">in</span><span class="p">);</span>
</pre></div>
</div>
<p>As you can see it is pretty similar to the Python API. When you call the <code class="docutils literal notranslate"><span class="pre">forward</span></code> method, you invoke the PyTorch JIT compiler, which will optimize and run your TorchScript code.</p>
</section>
<section id="compiling-with-torch-tensorrt-in-c">
<span id="compile-cpp"></span><h3>Compiling with Torch-TensorRT in C++<a class="headerlink" href="#compiling-with-torch-tensorrt-in-c" title="Permalink to this heading">¶</a></h3>
<p>We are also at the point where we can compile and optimize our module with Torch-TensorRT, but instead of in a JIT fashion we must do it ahead-of-time (AOT) i.e. before we start doing actual inference work
since it takes a bit of time to optimize the module, it would not make sense to do this every time you run the module or even the first time you run it.</p>
<p>With our module loaded, we can feed it into the Torch-TensorRT compiler. When we do so we must provide some information on the expected input size and also configure any additional settings.</p>
<div class="highlight-c++ notranslate"><div class="highlight"><pre><span></span><span class="cp">#include</span><span class="w"> </span><span class="cpf">"torch/script.h"</span>
<span class="cp">#include</span><span class="w"> </span><span class="cpf">"torch_tensorrt/torch_tensorrt.h"</span>
<span class="p">...</span>
<span class="w"> </span><span class="n">mod</span><span class="p">.</span><span class="n">to</span><span class="p">(</span><span class="n">at</span><span class="o">::</span><span class="n">kCUDA</span><span class="p">);</span>
<span class="w"> </span><span class="n">mod</span><span class="p">.</span><span class="n">eval</span><span class="p">();</span>
<span class="w"> </span><span class="n">std</span><span class="o">::</span><span class="n">vector</span><span class="o"><</span><span class="n">torch_tensorrt</span><span class="o">::</span><span class="n">core</span><span class="o">::</span><span class="n">ir</span><span class="o">::</span><span class="n">Input</span><span class="o">></span><span class="w"> </span><span class="n">inputs</span><span class="p">{</span><span class="n">torch_tensorrt</span><span class="o">::</span><span class="n">core</span><span class="o">::</span><span class="n">ir</span><span class="o">::</span><span class="n">Input</span><span class="p">({</span><span class="mi">1</span><span class="p">,</span><span class="w"> </span><span class="mi">3</span><span class="p">,</span><span class="w"> </span><span class="mi">224</span><span class="p">,</span><span class="w"> </span><span class="mi">224</span><span class="p">})};</span>
<span class="w"> </span><span class="n">torch_tensorrt</span><span class="o">::</span><span class="n">ts</span><span class="o">::</span><span class="n">CompileSpec</span><span class="w"> </span><span class="nf">cfg</span><span class="p">(</span><span class="n">inputs</span><span class="p">);</span>
<span class="w"> </span><span class="k">auto</span><span class="w"> </span><span class="n">trt_mod</span><span class="w"> </span><span class="o">=</span><span class="w"> </span><span class="n">torch_tensorrt</span><span class="o">::</span><span class="n">ts</span><span class="o">::</span><span class="n">compile</span><span class="p">(</span><span class="n">mod</span><span class="p">,</span><span class="w"> </span><span class="n">cfg</span><span class="p">);</span>
<span class="w"> </span><span class="k">auto</span><span class="w"> </span><span class="n">in</span><span class="w"> </span><span class="o">=</span><span class="w"> </span><span class="n">torch</span><span class="o">::</span><span class="n">randn</span><span class="p">({</span><span class="mi">1</span><span class="p">,</span><span class="w"> </span><span class="mi">3</span><span class="p">,</span><span class="w"> </span><span class="mi">224</span><span class="p">,</span><span class="w"> </span><span class="mi">224</span><span class="p">},</span><span class="w"> </span><span class="p">{</span><span class="n">torch</span><span class="o">::</span><span class="n">kCUDA</span><span class="p">});</span>
<span class="w"> </span><span class="k">auto</span><span class="w"> </span><span class="n">out</span><span class="w"> </span><span class="o">=</span><span class="w"> </span><span class="n">trt_mod</span><span class="p">.</span><span class="n">forward</span><span class="p">({</span><span class="n">in</span><span class="p">});</span>
</pre></div>
</div>
<p>That’s it! Now the graph runs primarily not with the JIT compiler but using TensorRT (though we execute the graph using the JIT runtime).</p>
<p>We can also set settings like operating precision to run in FP16.</p>
<div class="highlight-c++ notranslate"><div class="highlight"><pre><span></span><span class="cp">#include</span><span class="w"> </span><span class="cpf">"torch/script.h"</span>
<span class="cp">#include</span><span class="w"> </span><span class="cpf">"torch_tensorrt/torch_tensorrt.h"</span>
<span class="p">...</span>
<span class="w"> </span><span class="n">mod</span><span class="p">.</span><span class="n">to</span><span class="p">(</span><span class="n">at</span><span class="o">::</span><span class="n">kCUDA</span><span class="p">);</span>
<span class="w"> </span><span class="n">mod</span><span class="p">.</span><span class="n">eval</span><span class="p">();</span>
<span class="w"> </span><span class="k">auto</span><span class="w"> </span><span class="n">in</span><span class="w"> </span><span class="o">=</span><span class="w"> </span><span class="n">torch</span><span class="o">::</span><span class="n">randn</span><span class="p">({</span><span class="mi">1</span><span class="p">,</span><span class="w"> </span><span class="mi">3</span><span class="p">,</span><span class="w"> </span><span class="mi">224</span><span class="p">,</span><span class="w"> </span><span class="mi">224</span><span class="p">},</span><span class="w"> </span><span class="p">{</span><span class="n">torch</span><span class="o">::</span><span class="n">kCUDA</span><span class="p">}).</span><span class="n">to</span><span class="p">(</span><span class="n">torch</span><span class="o">::</span><span class="n">kHALF</span><span class="p">);</span>
<span class="w"> </span><span class="n">std</span><span class="o">::</span><span class="n">vector</span><span class="o"><</span><span class="n">torch_tensorrt</span><span class="o">::</span><span class="n">core</span><span class="o">::</span><span class="n">ir</span><span class="o">::</span><span class="n">Input</span><span class="o">></span><span class="w"> </span><span class="n">inputs</span><span class="p">{</span><span class="n">torch_tensorrt</span><span class="o">::</span><span class="n">core</span><span class="o">::</span><span class="n">ir</span><span class="o">::</span><span class="n">Input</span><span class="p">({</span><span class="mi">1</span><span class="p">,</span><span class="w"> </span><span class="mi">3</span><span class="p">,</span><span class="w"> </span><span class="mi">224</span><span class="p">,</span><span class="w"> </span><span class="mi">224</span><span class="p">})};</span>
<span class="w"> </span><span class="n">torch_tensorrt</span><span class="o">::</span><span class="n">ts</span><span class="o">::</span><span class="n">CompileSpec</span><span class="w"> </span><span class="nf">cfg</span><span class="p">(</span><span class="n">inputs</span><span class="p">);</span>
<span class="w"> </span><span class="n">cfg</span><span class="p">.</span><span class="n">enable_precisions</span><span class="p">.</span><span class="n">insert</span><span class="p">(</span><span class="n">torch</span><span class="o">::</span><span class="n">kHALF</span><span class="p">);</span>
<span class="w"> </span><span class="k">auto</span><span class="w"> </span><span class="n">trt_mod</span><span class="w"> </span><span class="o">=</span><span class="w"> </span><span class="n">torch_tensorrt</span><span class="o">::</span><span class="n">ts</span><span class="o">::</span><span class="n">compile</span><span class="p">(</span><span class="n">mod</span><span class="p">,</span><span class="w"> </span><span class="n">cfg</span><span class="p">);</span>
<span class="w"> </span><span class="k">auto</span><span class="w"> </span><span class="n">out</span><span class="w"> </span><span class="o">=</span><span class="w"> </span><span class="n">trt_mod</span><span class="p">.</span><span class="n">forward</span><span class="p">({</span><span class="n">in</span><span class="p">});</span>
</pre></div>
</div>
<p>And now we are running the module in FP16 precision. You can then save the module to load later.</p>
<div class="highlight-c++ notranslate"><div class="highlight"><pre><span></span><span class="n">trt_mod</span><span class="p">.</span><span class="n">save</span><span class="p">(</span><span class="s">"<PATH TO SAVED TRT/TS MOD>"</span><span class="p">)</span>
</pre></div>
</div>
<p>Torch-TensorRT compiled TorchScript modules are loaded in the same way as normal TorchScript module. Make sure your deployment application is linked against <code class="docutils literal notranslate"><span class="pre">libtorchtrt.so</span></code></p>
<div class="highlight-c++ notranslate"><div class="highlight"><pre><span></span><span class="cp">#include</span><span class="w"> </span><span class="cpf">"torch/script.h"</span>
<span class="cp">#include</span><span class="w"> </span><span class="cpf">"torch_tensorrt/torch_tensorrt.h"</span>
<span class="kt">int</span><span class="w"> </span><span class="nf">main</span><span class="p">(</span><span class="kt">int</span><span class="w"> </span><span class="n">argc</span><span class="p">,</span><span class="w"> </span><span class="k">const</span><span class="w"> </span><span class="kt">char</span><span class="o">*</span><span class="w"> </span><span class="n">argv</span><span class="p">[])</span><span class="w"> </span><span class="p">{</span>
<span class="w"> </span><span class="n">torch</span><span class="o">::</span><span class="n">jit</span><span class="o">::</span><span class="n">Module</span><span class="w"> </span><span class="k">module</span><span class="p">;</span>
<span class="w"> </span><span class="k">try</span><span class="w"> </span><span class="p">{</span>
<span class="w"> </span><span class="c1">// Deserialize the ScriptModule from a file using torch::jit::load().</span>
<span class="w"> </span><span class="k">module</span><span class="w"> </span><span class="o">=</span><span class="w"> </span><span class="n">torch</span><span class="o">::</span><span class="n">jit</span><span class="o">::</span><span class="n">load</span><span class="p">(</span><span class="s">"<PATH TO SAVED TRT/TS MOD>"</span><span class="p">);</span>
<span class="w"> </span><span class="p">}</span>
<span class="w"> </span><span class="k">catch</span><span class="w"> </span><span class="p">(</span><span class="k">const</span><span class="w"> </span><span class="n">c10</span><span class="o">::</span><span class="n">Error</span><span class="o">&</span><span class="w"> </span><span class="n">e</span><span class="p">)</span><span class="w"> </span><span class="p">{</span>
<span class="w"> </span><span class="n">std</span><span class="o">::</span><span class="n">cerr</span><span class="w"> </span><span class="o"><<</span><span class="w"> </span><span class="s">"error loading the model</span><span class="se">\n</span><span class="s">"</span><span class="p">;</span>
<span class="w"> </span><span class="k">return</span><span class="w"> </span><span class="mi">-1</span><span class="p">;</span>
<span class="w"> </span><span class="p">}</span>
<span class="w"> </span><span class="n">torch</span><span class="o">::</span><span class="n">Tensor</span><span class="w"> </span><span class="n">in</span><span class="w"> </span><span class="o">=</span><span class="w"> </span><span class="n">torch</span><span class="o">::</span><span class="n">randn</span><span class="p">({</span><span class="mi">1</span><span class="p">,</span><span class="w"> </span><span class="mi">1</span><span class="p">,</span><span class="w"> </span><span class="mi">32</span><span class="p">,</span><span class="w"> </span><span class="mi">32</span><span class="p">},</span><span class="w"> </span><span class="n">torch</span><span class="o">::</span><span class="n">kCUDA</span><span class="p">);</span>
<span class="w"> </span><span class="k">auto</span><span class="w"> </span><span class="n">out</span><span class="w"> </span><span class="o">=</span><span class="w"> </span><span class="n">mod</span><span class="p">.</span><span class="n">forward</span><span class="p">(</span><span class="n">in</span><span class="p">);</span>
<span class="w"> </span><span class="n">std</span><span class="o">::</span><span class="n">cout</span><span class="w"> </span><span class="o"><<</span><span class="w"> </span><span class="s">"ok</span><span class="se">\n</span><span class="s">"</span><span class="p">;</span>
<span class="p">}</span>
</pre></div>
</div>
<p>If you want to save the engine produced by Torch-TensorRT to use in a TensorRT application you can use the <code class="docutils literal notranslate"><span class="pre">ConvertGraphToTRTEngine</span></code> API.</p>
<div class="highlight-c++ notranslate"><div class="highlight"><pre><span></span><span class="cp">#include</span><span class="w"> </span><span class="cpf">"torch/script.h"</span>
<span class="cp">#include</span><span class="w"> </span><span class="cpf">"torch_tensorrt/torch_tensorrt.h"</span>
<span class="p">...</span>
<span class="w"> </span><span class="n">mod</span><span class="p">.</span><span class="n">to</span><span class="p">(</span><span class="n">at</span><span class="o">::</span><span class="n">kCUDA</span><span class="p">);</span>
<span class="w"> </span><span class="n">mod</span><span class="p">.</span><span class="n">eval</span><span class="p">();</span>
<span class="w"> </span><span class="k">auto</span><span class="w"> </span><span class="n">in</span><span class="w"> </span><span class="o">=</span><span class="w"> </span><span class="n">torch</span><span class="o">::</span><span class="n">randn</span><span class="p">({</span><span class="mi">1</span><span class="p">,</span><span class="w"> </span><span class="mi">3</span><span class="p">,</span><span class="w"> </span><span class="mi">224</span><span class="p">,</span><span class="w"> </span><span class="mi">224</span><span class="p">},</span><span class="w"> </span><span class="p">{</span><span class="n">torch</span><span class="o">::</span><span class="n">kCUDA</span><span class="p">}).</span><span class="n">to</span><span class="p">(</span><span class="n">torch</span><span class="o">::</span><span class="n">kHALF</span><span class="p">);</span>
<span class="w"> </span><span class="n">std</span><span class="o">::</span><span class="n">vector</span><span class="o"><</span><span class="n">torch_tensorrt</span><span class="o">::</span><span class="n">core</span><span class="o">::</span><span class="n">ir</span><span class="o">::</span><span class="n">Input</span><span class="o">></span><span class="w"> </span><span class="n">inputs</span><span class="p">{</span><span class="n">torch_tensorrt</span><span class="o">::</span><span class="n">core</span><span class="o">::</span><span class="n">ir</span><span class="o">::</span><span class="n">Input</span><span class="p">({</span><span class="mi">1</span><span class="p">,</span><span class="w"> </span><span class="mi">3</span><span class="p">,</span><span class="w"> </span><span class="mi">224</span><span class="p">,</span><span class="w"> </span><span class="mi">224</span><span class="p">})};</span>
<span class="w"> </span><span class="n">torch_tensorrt</span><span class="o">::</span><span class="n">ts</span><span class="o">::</span><span class="n">CompileSpec</span><span class="w"> </span><span class="nf">cfg</span><span class="p">(</span><span class="n">inputs</span><span class="p">);</span>
<span class="w"> </span><span class="n">cfg</span><span class="p">.</span><span class="n">enabled_precisions</span><span class="p">.</span><span class="n">insert</span><span class="p">(</span><span class="n">torch</span><span class="o">::</span><span class="n">kHALF</span><span class="p">);</span>
<span class="w"> </span><span class="k">auto</span><span class="w"> </span><span class="n">trt_mod</span><span class="w"> </span><span class="o">=</span><span class="w"> </span><span class="n">torch_tensorrt</span><span class="o">::</span><span class="n">ts</span><span class="o">::</span><span class="n">convert_method_to_trt_engine</span><span class="p">(</span><span class="n">mod</span><span class="p">,</span><span class="w"> </span><span class="s">"forward"</span><span class="p">,</span><span class="w"> </span><span class="n">cfg</span><span class="p">);</span>
<span class="w"> </span><span class="n">std</span><span class="o">::</span><span class="n">ofstream</span><span class="w"> </span><span class="n">out</span><span class="p">(</span><span class="s">"/tmp/engine_converted_from_jit.trt"</span><span class="p">);</span>
<span class="w"> </span><span class="n">out</span><span class="w"> </span><span class="o"><<</span><span class="w"> </span><span class="n">engine</span><span class="p">;</span>
<span class="w"> </span><span class="n">out</span><span class="p">.</span><span class="n">close</span><span class="p">();</span>
</pre></div>
</div>
</section>
<section id="under-the-hood">
<span id="id2"></span><h3>Under The Hood<a class="headerlink" href="#under-the-hood" title="Permalink to this heading">¶</a></h3>
<p>When a module is provided to Torch-TensorRT, the compiler starts by mapping a graph like you saw above to a graph like this:</p>
<div class="highlight-none notranslate"><div class="highlight"><pre><span></span>graph(%input.2 : Tensor):
%2 : Float(84, 10) = prim::Constant[value=<Tensor>]()
%3 : Float(120, 84) = prim::Constant[value=<Tensor>]()
%4 : Float(576, 120) = prim::Constant[value=<Tensor>]()
%5 : int = prim::Constant[value=-1]() # x.py:25:0
%6 : int[] = prim::Constant[value=annotate(List[int], [])]()
%7 : int[] = prim::Constant[value=[2, 2]]()
%8 : int[] = prim::Constant[value=[0, 0]]()
%9 : int[] = prim::Constant[value=[1, 1]]()
%10 : bool = prim::Constant[value=1]() # ~/.local/lib/python3.6/site-packages/torch/nn/modules/conv.py:346:0
%11 : int = prim::Constant[value=1]() # ~/.local/lib/python3.6/site-packages/torch/nn/functional.py:539:0
%12 : bool = prim::Constant[value=0]() # ~/.local/lib/python3.6/site-packages/torch/nn/functional.py:539:0
%self.classifier.fc3.bias : Float(10) = prim::Constant[value= 0.0464 0.0383 0.0678 0.0932 0.1045 -0.0805 -0.0435 -0.0818 0.0208 -0.0358 [ CUDAFloatType{10} ]]()
%self.classifier.fc2.bias : Float(84) = prim::Constant[value=<Tensor>]()
%self.classifier.fc1.bias : Float(120) = prim::Constant[value=<Tensor>]()
%self.feat.conv2.weight : Float(16, 6, 3, 3) = prim::Constant[value=<Tensor>]()
%self.feat.conv2.bias : Float(16) = prim::Constant[value=<Tensor>]()
%self.feat.conv1.weight : Float(6, 1, 3, 3) = prim::Constant[value=<Tensor>]()
%self.feat.conv1.bias : Float(6) = prim::Constant[value= 0.0530 -0.1691 0.2802 0.1502 0.1056 -0.1549 [ CUDAFloatType{6} ]]()
%input0.4 : Tensor = aten::_convolution(%input.2, %self.feat.conv1.weight, %self.feat.conv1.bias, %9, %8, %9, %12, %8, %11, %12, %12, %10) # ~/.local/lib/python3.6/site-packages/torch/nn/modules/conv.py:346:0
%input0.5 : Tensor = aten::relu(%input0.4) # ~/.local/lib/python3.6/site-packages/torch/nn/functional.py:1063:0
%input1.2 : Tensor = aten::max_pool2d(%input0.5, %7, %6, %8, %9, %12) # ~/.local/lib/python3.6/site-packages/torch/nn/functional.py:539:0
%input0.6 : Tensor = aten::_convolution(%input1.2, %self.feat.conv2.weight, %self.feat.conv2.bias, %9, %8, %9, %12, %8, %11, %12, %12, %10) # ~/.local/lib/python3.6/site-packages/torch/nn/modules/conv.py:346:0
%input2.1 : Tensor = aten::relu(%input0.6) # ~/.local/lib/python3.6/site-packages/torch/nn/functional.py:1063:0
%x.1 : Tensor = aten::max_pool2d(%input2.1, %7, %6, %8, %9, %12) # ~/.local/lib/python3.6/site-packages/torch/nn/functional.py:539:0
%input.1 : Tensor = aten::flatten(%x.1, %11, %5) # x.py:25:0
%27 : Tensor = aten::matmul(%input.1, %4)
%28 : Tensor = trt::const(%self.classifier.fc1.bias)
%29 : Tensor = aten::add_(%28, %27, %11)
%input0.2 : Tensor = aten::relu(%29) # ~/.local/lib/python3.6/site-packages/torch/nn/functional.py:1063:0
%31 : Tensor = aten::matmul(%input0.2, %3)
%32 : Tensor = trt::const(%self.classifier.fc2.bias)
%33 : Tensor = aten::add_(%32, %31, %11)
%input1.1 : Tensor = aten::relu(%33) # ~/.local/lib/python3.6/site-packages/torch/nn/functional.py:1063:0
%35 : Tensor = aten::matmul(%input1.1, %2)
%36 : Tensor = trt::const(%self.classifier.fc3.bias)
%37 : Tensor = aten::add_(%36, %35, %11)
return (%37)
(CompileGraph)
</pre></div>
</div>
<p>The graph has now been transformed from a collection of modules, each managing their own parameters into a single graph with the parameters inlined
into the graph and all of the operations laid out. Torch-TensorRT has also executed a number of optimizations and mappings to make the graph easier to translate to TensorRT.
From here the compiler can assemble the TensorRT engine by following the dataflow through the graph.</p>
<p>When the graph construction phase is complete, Torch-TensorRT produces a serialized TensorRT engine. From here depending on the API, this engine is returned
to the user or moves into the graph construction phase. Here Torch-TensorRT creates a JIT Module to execute the TensorRT engine which will be instantiated and managed
by the Torch-TensorRT runtime.</p>
<p>Here is the graph that you get back after compilation is complete:</p>
<div class="highlight-none notranslate"><div class="highlight"><pre><span></span>graph(%self_1 : __torch__.lenet, %input_0 : Tensor):
%1 : ...trt.Engine = prim::GetAttr[name="lenet"](%self_1)
%3 : Tensor[] = prim::ListConstruct(%input_0)
%4 : Tensor[] = trt::execute_engine(%3, %1)
%5 : Tensor = prim::ListUnpack(%4)
return (%5)
</pre></div>
</div>
<p>You can see the call where the engine is executed, after extracting the attribute containing the engine and constructing a list of inputs, then returns the tensors back to the user.</p>
</section>
<section id="working-with-unsupported-operators">
<span id="unsupported-ops"></span><h3>Working with Unsupported Operators<a class="headerlink" href="#working-with-unsupported-operators" title="Permalink to this heading">¶</a></h3>
<p>Torch-TensorRT is a new library and the PyTorch operator library is quite large, so there will be ops that aren’t supported natively by the compiler. You can either use the composition techniques
shown above to make modules are fully Torch-TensorRT supported and ones that are not and stitch the modules together in the deployment application or you can register converters for missing ops.</p>
<blockquote>
<div><p>You can check support without going through the full compilation pipeline using the <code class="docutils literal notranslate"><span class="pre">torch_tensorrt::CheckMethodOperatorSupport(const</span> <span class="pre">torch::jit::Module&</span> <span class="pre">module,</span> <span class="pre">std::string</span> <span class="pre">method_name)</span></code> api
to see what operators are not supported. <code class="docutils literal notranslate"><span class="pre">torchtrtc</span></code> automatically checks modules with this method before starting compilation and will print out a list of operators that are not supported.</p>
</div></blockquote>
<section id="registering-custom-converters">
<span id="custom-converters"></span><h4>Registering Custom Converters<a class="headerlink" href="#registering-custom-converters" title="Permalink to this heading">¶</a></h4>
<p>Operations are mapped to TensorRT through the use of modular converters, a function that takes a node from a the JIT graph and produces an equivalent layer or subgraph in TensorRT.
Torch-TensorRT ships with a library of these converters stored in a registry, that will be executed depending on the node being parsed. For instance a <code class="docutils literal notranslate"><span class="pre">aten::relu(%input0.4)</span></code> instruction will trigger
the relu converter to be run on it, producing an activation layer in the TensorRT graph. But since this library is not exhaustive you may need to write your own to get Torch-TensorRT
to support your module.</p>
<p>Shipped with the Torch-TensorRT distribution are the internal core API headers. You can therefore access the converter registry and add a converter for the op you need.</p>
<p>For example, if we try to compile a graph with a build of Torch-TensorRT that doesn’t support the flatten operation (<code class="docutils literal notranslate"><span class="pre">aten::flatten</span></code>) you may see this error:</p>
<div class="highlight-none notranslate"><div class="highlight"><pre><span></span>terminate called after throwing an instance of 'torch_tensorrt::Error'
what(): [enforce fail at core/conversion/conversion.cpp:109] Expected converter to be true but got false
Unable to convert node: %input.1 : Tensor = aten::flatten(%x.1, %11, %5) # x.py:25:0 (conversion.AddLayer)
Schema: aten::flatten.using_ints(Tensor self, int start_dim=0, int end_dim=-1) -> (Tensor)
Converter for aten::flatten requested, but no such converter was found.
If you need a converter for this operator, you can try implementing one yourself
or request a converter: https://www.github.com/NVIDIA/Torch-TensorRT/issues
</pre></div>
</div>
<p>We can register a converter for this operator in our application. All of the tools required to build a converter can be imported by including <code class="docutils literal notranslate"><span class="pre">torch_tensorrt/core/conversion/converters/converters.h</span></code>.
We start by creating an instance of the self-registering class <code class="docutils literal notranslate"><span class="pre">torch_tensorrt::core::conversion::converters::RegisterNodeConversionPatterns()</span></code> which will register converters
in the global converter registry, associating a function schema like <code class="docutils literal notranslate"><span class="pre">aten::flatten.using_ints(Tensor</span> <span class="pre">self,</span> <span class="pre">int</span> <span class="pre">start_dim=0,</span> <span class="pre">int</span> <span class="pre">end_dim=-1)</span> <span class="pre">-></span> <span class="pre">(Tensor)</span></code> with a lambda that
will take the state of the conversion, the node/operation in question to convert and all of the inputs to the node and produces as a side effect a new layer in the TensorRT network.
Arguments are passed as a vector of inspectable unions of TensorRT <code class="docutils literal notranslate"><span class="pre">ITensors</span></code> and Torch <code class="docutils literal notranslate"><span class="pre">IValues</span></code> in the order arguments are listed in the schema.</p>
<p>Below is a implementation of a <code class="docutils literal notranslate"><span class="pre">aten::flatten</span></code> converter that we can use in our application. You have full access to the Torch and TensorRT libraries in the converter implementation. So
for example we can quickly get the output size by just running the operation in PyTorch instead of implementing the full calculation outself like we do below for this flatten converter.</p>
<div class="highlight-c++ notranslate"><div class="highlight"><pre><span></span><span class="cp">#include</span><span class="w"> </span><span class="cpf">"torch/script.h"</span>
<span class="cp">#include</span><span class="w"> </span><span class="cpf">"torch_tensorrt/torch_tensorrt.h"</span>
<span class="cp">#include</span><span class="w"> </span><span class="cpf">"torch_tensorrt/core/conversion/converters/converters.h"</span>
<span class="k">static</span><span class="w"> </span><span class="k">auto</span><span class="w"> </span><span class="n">flatten_converter</span><span class="w"> </span><span class="o">=</span><span class="w"> </span><span class="n">torch_tensorrt</span><span class="o">::</span><span class="n">core</span><span class="o">::</span><span class="n">conversion</span><span class="o">::</span><span class="n">converters</span><span class="o">::</span><span class="n">RegisterNodeConversionPatterns</span><span class="p">()</span>
<span class="w"> </span><span class="p">.</span><span class="n">pattern</span><span class="p">({</span>
<span class="w"> </span><span class="s">"aten::flatten.using_ints(Tensor self, int start_dim=0, int end_dim=-1) -> (Tensor)"</span><span class="p">,</span>
<span class="w"> </span><span class="p">[](</span><span class="n">torch_tensorrt</span><span class="o">::</span><span class="n">core</span><span class="o">::</span><span class="n">conversion</span><span class="o">::</span><span class="n">ConversionCtx</span><span class="o">*</span><span class="w"> </span><span class="n">ctx</span><span class="p">,</span>
<span class="w"> </span><span class="k">const</span><span class="w"> </span><span class="n">torch</span><span class="o">::</span><span class="n">jit</span><span class="o">::</span><span class="n">Node</span><span class="o">*</span><span class="w"> </span><span class="n">n</span><span class="p">,</span>
<span class="w"> </span><span class="n">torch_tensorrt</span><span class="o">::</span><span class="n">core</span><span class="o">::</span><span class="n">conversion</span><span class="o">::</span><span class="n">converters</span><span class="o">::</span><span class="n">args</span><span class="o">&</span><span class="w"> </span><span class="n">args</span><span class="p">)</span><span class="w"> </span><span class="o">-></span><span class="w"> </span><span class="kt">bool</span><span class="w"> </span><span class="p">{</span>
<span class="w"> </span><span class="k">auto</span><span class="w"> </span><span class="n">in</span><span class="w"> </span><span class="o">=</span><span class="w"> </span><span class="n">args</span><span class="p">[</span><span class="mi">0</span><span class="p">].</span><span class="n">ITensor</span><span class="p">();</span>
<span class="w"> </span><span class="k">auto</span><span class="w"> </span><span class="n">start_dim</span><span class="w"> </span><span class="o">=</span><span class="w"> </span><span class="n">args</span><span class="p">[</span><span class="mi">1</span><span class="p">].</span><span class="n">unwrapToInt</span><span class="p">();</span>
<span class="w"> </span><span class="k">auto</span><span class="w"> </span><span class="n">end_dim</span><span class="w"> </span><span class="o">=</span><span class="w"> </span><span class="n">args</span><span class="p">[</span><span class="mi">2</span><span class="p">].</span><span class="n">unwrapToInt</span><span class="p">();</span>
<span class="w"> </span><span class="k">auto</span><span class="w"> </span><span class="n">in_shape</span><span class="w"> </span><span class="o">=</span><span class="w"> </span><span class="n">torch_tensorrt</span><span class="o">::</span><span class="n">core</span><span class="o">::</span><span class="n">util</span><span class="o">::</span><span class="n">toVec</span><span class="p">(</span><span class="n">in</span><span class="o">-></span><span class="n">getDimensions</span><span class="p">());</span>
<span class="w"> </span><span class="k">auto</span><span class="w"> </span><span class="n">out_shape</span><span class="w"> </span><span class="o">=</span><span class="w"> </span><span class="n">torch</span><span class="o">::</span><span class="n">flatten</span><span class="p">(</span><span class="n">torch</span><span class="o">::</span><span class="n">rand</span><span class="p">(</span><span class="n">in_shape</span><span class="p">),</span><span class="w"> </span><span class="n">start_dim</span><span class="p">,</span><span class="w"> </span><span class="n">end_dim</span><span class="p">).</span><span class="n">sizes</span><span class="p">();</span>
<span class="w"> </span><span class="k">auto</span><span class="w"> </span><span class="n">shuffle</span><span class="w"> </span><span class="o">=</span><span class="w"> </span><span class="n">ctx</span><span class="o">-></span><span class="n">net</span><span class="o">-></span><span class="n">addShuffle</span><span class="p">(</span><span class="o">*</span><span class="n">in</span><span class="p">);</span>
<span class="w"> </span><span class="n">shuffle</span><span class="o">-></span><span class="n">setReshapeDimensions</span><span class="p">(</span><span class="n">torch_tensorrt</span><span class="o">::</span><span class="n">core</span><span class="o">::</span><span class="n">util</span><span class="o">::</span><span class="n">toDims</span><span class="p">(</span><span class="n">out_shape</span><span class="p">));</span>
<span class="w"> </span><span class="n">shuffle</span><span class="o">-></span><span class="n">setName</span><span class="p">(</span><span class="n">torch_tensorrt</span><span class="o">::</span><span class="n">core</span><span class="o">::</span><span class="n">util</span><span class="o">::</span><span class="n">node_info</span><span class="p">(</span><span class="n">n</span><span class="p">).</span><span class="n">c_str</span><span class="p">());</span>
<span class="w"> </span><span class="k">auto</span><span class="w"> </span><span class="n">out_tensor</span><span class="w"> </span><span class="o">=</span><span class="w"> </span><span class="n">ctx</span><span class="o">-></span><span class="n">AssociateValueAndTensor</span><span class="p">(</span><span class="n">n</span><span class="o">-></span><span class="n">outputs</span><span class="p">()[</span><span class="mi">0</span><span class="p">],</span><span class="w"> </span><span class="n">shuffle</span><span class="o">-></span><span class="n">getOutput</span><span class="p">(</span><span class="mi">0</span><span class="p">));</span>
<span class="w"> </span><span class="k">return</span><span class="w"> </span><span class="nb">true</span><span class="p">;</span>
<span class="w"> </span><span class="p">}</span>
<span class="w"> </span><span class="p">});</span>
<span class="kt">int</span><span class="w"> </span><span class="nf">main</span><span class="p">()</span><span class="w"> </span><span class="p">{</span>
<span class="w"> </span><span class="p">...</span>
</pre></div>
</div>
<p>To use this converter in Python, it is recommended to use PyTorch’s <a class="reference external" href="https://pytorch.org/tutorials/advanced/cpp_extension.html#custom-c-and-cuda-extensions">C++ / CUDA Extension</a>
template to wrap your library of converters into a <code class="docutils literal notranslate"><span class="pre">.so</span></code> that you can load with <code class="docutils literal notranslate"><span class="pre">ctypes.CDLL()</span></code> in your Python application.</p>
<p>You can find more information on all the details of writing converters in the contributors documentation (<span class="xref std std-ref">writing_converters</span>).
If you find yourself with a large library of converter implementations, do consider upstreaming them, PRs are welcome and it would be great for the community to benefit as well.</p>
</section>
</section>
</section>
</section>
</article>
</div>
<footer>
<div class="rst-footer-buttons" role="navigation" aria-label="footer navigation">
<a href="ptq.html" class="btn btn-neutral float-right" title="Post Training Quantization (PTQ)" accesskey="n" rel="next">Next <img src="../_static/images/chevron-right-orange.svg" class="next-page"></a>
<a href="getting_started_with_python_api.html" class="btn btn-neutral" title="Using Torch-TensorRT in Python" accesskey="p" rel="prev"><img src="../_static/images/chevron-right-orange.svg" class="previous-page"> Previous</a>
</div>
<hr>
<div role="contentinfo">
<p>
© Copyright 2024, NVIDIA Corporation.
</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>
</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="#">Using Torch-TensorRT in C++</a><ul>
<li><a class="reference internal" href="#id1">Using Torch-TensorRT in C++</a><ul>
<li><a class="reference internal" href="#torch-tensorrt-quickstart-compiling-torchscript-modules-with-torchtrtc">[Torch-TensorRT Quickstart] Compiling TorchScript Modules with <code class="docutils literal notranslate"><span class="pre">torchtrtc</span></code></a></li>
<li><a class="reference internal" href="#working-with-torchscript-in-c">Working with TorchScript in C++</a></li>
<li><a class="reference internal" href="#compiling-with-torch-tensorrt-in-c">Compiling with Torch-TensorRT in C++</a></li>
<li><a class="reference internal" href="#under-the-hood">Under The Hood</a></li>
<li><a class="reference internal" href="#working-with-unsupported-operators">Working with Unsupported Operators</a><ul>
<li><a class="reference internal" href="#registering-custom-converters">Registering Custom Converters</a></li>
</ul>
</li>
</ul>
</li>
</ul>
</li>
</ul>
</div>
</div>
</div>
</section>
</div>
<script type="text/javascript" id="documentation_options" data-url_root="../" src="../_static/documentation_options.js"></script>
<script data-url_root="../" id="documentation_options" src="../_static/documentation_options.js"></script>
<script src="../_static/jquery.js"></script>
<script src="../_static/underscore.js"></script>
<script src="../_static/_sphinx_javascript_frameworks_compat.js"></script>
<script src="../_static/doctools.js"></script>
<script src="../_static/collapsible-lists/js/CollapsibleLists.compressed.js"></script>
<script src="../_static/collapsible-lists/js/apply-collapsible-lists.js"></script>
<script crossorigin="anonymous" integrity="sha256-Ae2Vz/4ePdIu6ZyI/5ZGsYnb+m0JlOmKPjt6XZ9JJkA=" src="https://cdnjs.cloudflare.com/ajax/libs/require.js/2.3.4/require.min.js"></script>
<script type="text/javascript" src="../_static/js/vendor/popper.min.js"></script>
<script type="text/javascript" src="../_static/js/vendor/bootstrap.min.js"></script>
<script src="https://cdnjs.cloudflare.com/ajax/libs/list.js/1.5.0/list.min.js"></script>
<script type="text/javascript" src="../_static/js/theme.js"></script>
<script type="text/javascript">
jQuery(function () {
SphinxRtdTheme.Navigation.enable(true);
});
</script>
<!-- Begin Footer -->
<div class="container-fluid docs-tutorials-resources" id="docs-tutorials-resources">
<div class="container">
<div class="row">
<div class="col-md-4 text-center">
<h2>Docs</h2>
<p>Access comprehensive developer documentation for PyTorch</p>
<a class="with-right-arrow" href="https://pytorch.org/docs/stable/index.html">View Docs</a>
</div>
<div class="col-md-4 text-center">
<h2>Tutorials</h2>
<p>Get in-depth tutorials for beginners and advanced developers</p>
<a class="with-right-arrow" href="https://pytorch.org/tutorials">View Tutorials</a>
</div>
<div class="col-md-4 text-center">
<h2>Resources</h2>
<p>Find development resources and get your questions answered</p>
<a class="with-right-arrow" href="https://pytorch.org/resources">View Resources</a>
</div>
</div>
</div>
</div>
<footer class="site-footer">
<div class="container footer-container">
<div class="footer-logo-wrapper">
<a href="https://pytorch.org/" class="footer-logo"></a>
</div>
<div class="footer-links-wrapper">
<div class="footer-links-col">
<ul>
<li class="list-title"><a href="https://pytorch.org/">PyTorch</a></li>
<li><a href="https://pytorch.org/get-started">Get Started</a></li>
<li><a href="https://pytorch.org/features">Features</a></li>
<li><a href="https://pytorch.org/ecosystem">Ecosystem</a></li>
<li><a href="https://pytorch.org/blog/">Blog</a></li>
<li><a href="https://github.com/pytorch/pytorch/blob/master/CONTRIBUTING.md">Contributing</a></li>
</ul>
</div>
<div class="footer-links-col">
<ul>
<li class="list-title"><a href="https://pytorch.org/resources">Resources</a></li>
<li><a href="https://pytorch.org/tutorials">Tutorials</a></li>
<li><a href="https://pytorch.org/docs/stable/index.html">Docs</a></li>
<li><a href="https://discuss.pytorch.org" target="_blank">Discuss</a></li>
<li><a href="https://github.com/pytorch/pytorch/issues" target="_blank">Github Issues</a></li>
<li><a href="https://pytorch.org/assets/brand-guidelines/PyTorch-Brand-Guidelines.pdf" target="_blank">Brand Guidelines</a></li>
</ul>
</div>
<div class="footer-links-col">
<ul>
<li class="list-title">Stay up to date</li>
<li><a href="https://www.facebook.com/pytorch" target="_blank">Facebook</a></li>
<li><a href="https://twitter.com/pytorch" target="_blank">Twitter</a></li>
<li><a href="https://www.youtube.com/pytorch" target="_blank">YouTube</a></li>
<li><a href="https://www.linkedin.com/company/pytorch" target="_blank">LinkedIn</a></li>
</ul>
</div>
<div class="footer-links-col">
<ul>
<li class="list-title">PyTorch Podcasts</li>
<li><a href="https://open.spotify.com/show/6UzHKeiy368jKfQMKKvJY5" target="_blank">Spotify</a></li>
<li><a href="https://podcasts.apple.com/us/podcast/pytorch-developer-podcast/id1566080008" target="_blank">Apple</a></li>
<li><a href="https://www.google.com/podcasts?feed=aHR0cHM6Ly9mZWVkcy5zaW1wbGVjYXN0LmNvbS9PQjVGa0lsOA%3D%3D" target="_blank">Google</a></li>
<li><a href="https://music.amazon.com/podcasts/7a4e6f0e-26c2-49e9-a478-41bd244197d0/PyTorch-Developer-Podcast?" target="_blank">Amazon</a></li>
</ul>
</div>
</div>
<div class="privacy-policy">
<ul>
<li class="privacy-policy-links"><a href="https://www.linuxfoundation.org/terms/" target="_blank">Terms</a></li>
<li class="privacy-policy-links">|</li>
<li class="privacy-policy-links"><a href="https://www.linuxfoundation.org/privacy-policy/" target="_blank">Privacy</a></li>
</ul>
</div>
<div class="copyright">
<p>© Copyright The Linux Foundation. The PyTorch Foundation is a project of The Linux Foundation.
For web site terms of use, trademark policy and other policies applicable to The PyTorch Foundation please see
<a href="https://www.linuxfoundation.org/policies/">www.linuxfoundation.org/policies/</a>. The PyTorch Foundation supports the PyTorch open source
project, which has been established as PyTorch Project a Series of LF Projects, LLC. For policies applicable to the PyTorch Project a Series of LF Projects, LLC,
please see <a href="https://www.lfprojects.org/policies/">www.lfprojects.org/policies/</a>.</p>
</div>
</div>
</footer>
<div class="cookie-banner-wrapper">
<div class="container">
<p class="gdpr-notice">To analyze traffic and optimize your experience, we serve cookies on this site. By clicking or navigating, you agree to allow our usage of cookies. As the current maintainers of this site, Facebook’s Cookies Policy applies. Learn more, including about available controls: <a href="https://www.facebook.com/policies/cookies/">Cookies Policy</a>.</p>
<img class="close-button" src="../_static/images/pytorch-x.svg">
</div>
</div>
<!-- End Footer -->
<!-- Begin Mobile Menu -->
<div class="mobile-main-menu">
<div class="container-fluid">
<div class="container">
<div class="mobile-main-menu-header-container">
<a class="header-logo" href="https://pytorch.org/" aria-label="PyTorch"></a>
<a class="main-menu-close-button" href="#" data-behavior="close-mobile-menu"></a>
</div>
</div>
</div>
<div class="mobile-main-menu-links-container">
<div class="main-menu">
<ul>
<li class="resources-mobile-menu-title">
<a>Learn</a>
</li>
<ul class="resources-mobile-menu-items">
<li>
<a href="https://pytorch.org/get-started">Get Started</a>
</li>
<li>