-
Notifications
You must be signed in to change notification settings - Fork 362
/
Copy pathgetting_started.html
1435 lines (1413 loc) · 77.4 KB
/
getting_started.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>
<html>
<head>
<meta charset="utf-8"/>
<meta content="width=device-width,initial-scale=1" name="viewport"/>
<meta content="ie=edge" http-equiv="x-ua-compatible"/>
<meta content="Copy to clipboard" name="lang:clipboard.copy"/>
<meta content="Copied to clipboard" name="lang:clipboard.copied"/>
<meta content="en" name="lang:search.language"/>
<meta content="True" name="lang:search.pipeline.stopwords"/>
<meta content="True" name="lang:search.pipeline.trimmer"/>
<meta content="No matching documents" name="lang:search.result.none"/>
<meta content="1 matching document" name="lang:search.result.one"/>
<meta content="# matching documents" name="lang:search.result.other"/>
<meta content="[\s\-]+" name="lang:search.tokenizer"/>
<link crossorigin="" href="https://fonts.gstatic.com/" rel="preconnect"/>
<link href="https://fonts.googleapis.com/css?family=Roboto+Mono:400,500,700|Roboto:300,400,400i,700&display=fallback" rel="stylesheet"/>
<style>
body,
input {
font-family: "Roboto", "Helvetica Neue", Helvetica, Arial, sans-serif
}
code,
kbd,
pre {
font-family: "Roboto Mono", "Courier New", Courier, monospace
}
</style>
<link href="../_static/stylesheets/application.css" rel="stylesheet"/>
<link href="../_static/stylesheets/application-palette.css" rel="stylesheet"/>
<link href="../_static/stylesheets/application-fixes.css" rel="stylesheet"/>
<link href="../_static/fonts/material-icons.css" rel="stylesheet"/>
<meta content="84bd00" name="theme-color"/>
<script src="../_static/javascripts/modernizr.js">
</script>
<title>
Getting Started — TRTorch 0.0.2 documentation
</title>
<link href="../_static/material.css" rel="stylesheet" type="text/css"/>
<link href="../_static/pygments.css" rel="stylesheet" type="text/css"/>
<link href="../_static/collapsible-lists/css/tree_view.css" rel="stylesheet" type="text/css"/>
<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/doctools.js">
</script>
<script src="../_static/language_data.js">
</script>
<script src="../_static/collapsible-lists/js/CollapsibleLists.compressed.js">
</script>
<script src="../_static/collapsible-lists/js/apply-collapsible-lists.js">
</script>
<link href="../genindex.html" rel="index" title="Index"/>
<link href="../search.html" rel="search" title="Search"/>
<link href="ptq.html" rel="next" title="Post Training Quantization (PTQ)"/>
<link href="installation.html" rel="prev" title="Installation"/>
</head>
<body data-md-color-accent="light-green" data-md-color-primary="light-green" dir="ltr">
<svg class="md-svg">
<defs data-children-count="0">
<svg height="448" id="__github" viewbox="0 0 416 448" width="416" xmlns="http://www.w3.org/2000/svg">
<path d="M160 304q0 10-3.125 20.5t-10.75 19T128 352t-18.125-8.5-10.75-19T96 304t3.125-20.5 10.75-19T128 256t18.125 8.5 10.75 19T160 304zm160 0q0 10-3.125 20.5t-10.75 19T288 352t-18.125-8.5-10.75-19T256 304t3.125-20.5 10.75-19T288 256t18.125 8.5 10.75 19T320 304zm40 0q0-30-17.25-51T296 232q-10.25 0-48.75 5.25Q229.5 240 208 240t-39.25-2.75Q130.75 232 120 232q-29.5 0-46.75 21T56 304q0 22 8 38.375t20.25 25.75 30.5 15 35 7.375 37.25 1.75h42q20.5 0 37.25-1.75t35-7.375 30.5-15 20.25-25.75T360 304zm56-44q0 51.75-15.25 82.75-9.5 19.25-26.375 33.25t-35.25 21.5-42.5 11.875-42.875 5.5T212 416q-19.5 0-35.5-.75t-36.875-3.125-38.125-7.5-34.25-12.875T37 371.5t-21.5-28.75Q0 312 0 260q0-59.25 34-99-6.75-20.5-6.75-42.5 0-29 12.75-54.5 27 0 47.5 9.875t47.25 30.875Q171.5 96 212 96q37 0 70 8 26.25-20.5 46.75-30.25T376 64q12.75 25.5 12.75 54.5 0 21.75-6.75 42 34 40 34 99.5z" fill="currentColor">
</path>
</svg>
</defs>
</svg>
<input class="md-toggle" data-md-toggle="drawer" id="__drawer" type="checkbox"/>
<input class="md-toggle" data-md-toggle="search" id="__search" type="checkbox"/>
<label class="md-overlay" data-md-component="overlay" for="__drawer">
</label>
<a class="md-skip" href="#tutorials/getting_started" tabindex="1">
Skip to content
</a>
<header class="md-header" data-md-component="header">
<nav class="md-header-nav md-grid">
<div class="md-flex navheader">
<div class="md-flex__cell md-flex__cell--shrink">
<a class="md-header-nav__button md-logo" href="../index.html" title="TRTorch 0.0.2 documentation">
<i class="md-icon">
</i>
</a>
</div>
<div class="md-flex__cell md-flex__cell--shrink">
<label class="md-icon md-icon--menu md-header-nav__button" for="__drawer">
</label>
</div>
<div class="md-flex__cell md-flex__cell--stretch">
<div class="md-flex__ellipsis md-header-nav__title" data-md-component="title">
<span class="md-header-nav__topic">
TRTorch
</span>
<span class="md-header-nav__topic">
Getting Started
</span>
</div>
</div>
<div class="md-flex__cell md-flex__cell--shrink">
<label class="md-icon md-icon--search md-header-nav__button" for="__search">
</label>
<div class="md-search" data-md-component="search" role="dialog">
<label class="md-search__overlay" for="__search">
</label>
<div class="md-search__inner" role="search">
<form action="../search.html" class="md-search__form" method="GET" name="search">
<input autocapitalize="off" autocomplete="off" class="md-search__input" data-md-component="query" data-md-state="active" name="q" placeholder="Search" spellcheck="false" type="text"/>
<label class="md-icon md-search__icon" for="__search">
</label>
<button class="md-icon md-search__icon" data-md-component="reset" tabindex="-1" type="reset">
</button>
</form>
<div class="md-search__output">
<div class="md-search__scrollwrap" data-md-scrollfix="">
<div class="md-search-result" data-md-component="result">
<div class="md-search-result__meta">
Type to start searching
</div>
<ol class="md-search-result__list">
</ol>
</div>
</div>
</div>
</div>
</div>
</div>
<div class="md-flex__cell md-flex__cell--shrink">
<div class="md-header-nav__source">
<a class="md-source" data-md-source="github" href="https://github.com/nvidia/TRTorch/" title="Go to repository">
<div class="md-source__icon">
<svg height="28" viewbox="0 0 24 24" width="28" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink">
<use height="24" width="24" xlink:href="#__github">
</use>
</svg>
</div>
<div class="md-source__repository">
TRTorch
</div>
</a>
</div>
</div>
<div class="md-flex__cell md-flex__cell--shrink dropdown">
<button class="dropdownbutton">
Versions
</button>
<div class="dropdown-content md-hero">
<a href="https://nvidia.github.io/TRTorch/" title="master">
master
</a>
<a href="https://nvidia.github.io/TRTorch/v0.0.2/" title="v0.0.2">
v0.0.2
</a>
<a href="https://nvidia.github.io/TRTorch/v0.0.1/" title="v0.0.1">
v0.0.1
</a>
</div>
</div>
</div>
</nav>
</header>
<div class="md-container">
<nav class="md-tabs" data-md-component="tabs">
<div class="md-tabs__inner md-grid">
<ul class="md-tabs__list">
<li class="md-tabs__item">
<a class="md-tabs__link" href="../index.html">
TRTorch 0.0.2 documentation
</a>
</li>
</ul>
</div>
</nav>
<main class="md-main">
<div class="md-main__inner md-grid" data-md-component="container">
<div class="md-sidebar md-sidebar--primary" data-md-component="navigation">
<div class="md-sidebar__scrollwrap">
<div class="md-sidebar__inner">
<nav class="md-nav md-nav--primary" data-md-level="0">
<label class="md-nav__title md-nav__title--site" for="__drawer">
<a class="md-nav__button md-logo" href="../index.html" title="TRTorch 0.0.2 documentation">
<i class="md-icon">
</i>
</a>
<a href="../index.html" title="TRTorch 0.0.2 documentation">
TRTorch
</a>
</label>
<div class="md-nav__source">
<a class="md-source" data-md-source="github" href="https://github.com/nvidia/TRTorch/" title="Go to repository">
<div class="md-source__icon">
<svg height="28" viewbox="0 0 24 24" width="28" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink">
<use height="24" width="24" xlink:href="#__github">
</use>
</svg>
</div>
<div class="md-source__repository">
TRTorch
</div>
</a>
</div>
<ul class="md-nav__list">
<li class="md-nav__item">
<span class="md-nav__link caption">
<span class="caption-text">
Getting Started
</span>
</span>
</li>
<li class="md-nav__item">
<a class="md-nav__link" href="installation.html">
Installation
</a>
<ul class="md-nav__list">
<li class="md-nav__item">
<a class="md-nav__link" href="installation.html#dependencies">
Dependencies
</a>
</li>
<li class="md-nav__item">
<a class="md-nav__link" href="installation.html#dependencies-for-compilation">
Dependencies for Compilation
</a>
</li>
<li class="md-nav__item">
<a class="md-nav__link" href="installation.html#building-using-cudnn-tensorrt-tarball-distributions">
<strong>
Building using cuDNN & TensorRT tarball distributions
</strong>
</a>
</li>
<li class="md-nav__item">
<a class="md-nav__link" href="installation.html#building-using-locally-installed-cudnn-tensorrt">
<strong>
Building using locally installed cuDNN & TensorRT
</strong>
</a>
</li>
</ul>
</li>
<li class="md-nav__item">
<input class="md-toggle md-nav__toggle" data-md-toggle="toc" id="__toc" type="checkbox"/>
<label class="md-nav__link md-nav__link--active" for="__toc">
Getting Started
</label>
<a class="md-nav__link md-nav__link--active" href="#">
Getting Started
</a>
<nav class="md-nav md-nav--secondary">
<label class="md-nav__title" for="__toc">
Contents
</label>
<ul class="md-nav__list" data-md-scrollfix="">
<li class="md-nav__item">
<a class="md-nav__link" href="#tutorials-getting-started--page-root">
Getting Started
</a>
<nav class="md-nav">
<ul class="md-nav__list">
<li class="md-nav__item">
<a class="md-nav__link" href="#creating-a-torchscript-module">
Creating a TorchScript Module
</a>
</li>
<li class="md-nav__item">
<a class="md-nav__link" href="#working-with-torchscript-in-python">
Working with TorchScript in Python
</a>
</li>
<li class="md-nav__item">
<a class="md-nav__link" href="#compiling-with-trtorch-in-python">
Compiling with TRTorch in Python
</a>
</li>
<li class="md-nav__item">
<a class="md-nav__link" href="#working-with-torchscript-in-c">
Working with TorchScript in C++
</a>
</li>
<li class="md-nav__item">
<a class="md-nav__link" href="#compiling-with-trtorch-in-c">
Compiling with TRTorch in C++
</a>
</li>
<li class="md-nav__item">
<a class="md-nav__link" href="#under-the-hood">
Under The Hood
</a>
</li>
<li class="md-nav__item">
<a class="md-nav__link" href="#working-with-unsupported-operators">
Working with Unsupported Operators
</a>
<nav class="md-nav">
<ul class="md-nav__list">
<li class="md-nav__item">
<a class="md-nav__link" href="#registering-custom-converters">
Registering Custom Converters
</a>
</li>
</ul>
</nav>
</li>
</ul>
</nav>
</li>
<li class="md-nav__item">
<a class="md-nav__extra_link" href="../_sources/tutorials/getting_started.rst.txt">
Show Source
</a>
</li>
</ul>
</nav>
<ul class="md-nav__list">
<li class="md-nav__item">
<a class="md-nav__link" href="#creating-a-torchscript-module">
Creating a TorchScript Module
</a>
</li>
<li class="md-nav__item">
<a class="md-nav__link" href="#working-with-torchscript-in-python">
Working with TorchScript in Python
</a>
</li>
<li class="md-nav__item">
<a class="md-nav__link" href="#compiling-with-trtorch-in-python">
Compiling with TRTorch in Python
</a>
</li>
<li class="md-nav__item">
<a class="md-nav__link" href="#working-with-torchscript-in-c">
Working with TorchScript in C++
</a>
</li>
<li class="md-nav__item">
<a class="md-nav__link" href="#compiling-with-trtorch-in-c">
Compiling with TRTorch in C++
</a>
</li>
<li class="md-nav__item">
<a class="md-nav__link" href="#under-the-hood">
Under The Hood
</a>
</li>
<li class="md-nav__item">
<a class="md-nav__link" href="#working-with-unsupported-operators">
Working with Unsupported Operators
</a>
</li>
</ul>
</li>
<li class="md-nav__item">
<a class="md-nav__link" href="ptq.html">
Post Training Quantization (PTQ)
</a>
<ul class="md-nav__list">
<li class="md-nav__item">
<a class="md-nav__link" href="ptq.html#how-to-create-your-own-ptq-application">
How to create your own PTQ application
</a>
</li>
</ul>
</li>
<li class="md-nav__item">
<span class="md-nav__link caption">
<span class="caption-text">
Contributor Documentation
</span>
</span>
</li>
<li class="md-nav__item">
<a class="md-nav__link" href="../contributors/system_overview.html">
System Overview
</a>
<ul class="md-nav__list">
<li class="md-nav__item">
<a class="md-nav__link" href="../contributors/system_overview.html#compiler-phases">
Compiler Phases
</a>
</li>
</ul>
</li>
<li class="md-nav__item">
<a class="md-nav__link" href="../contributors/writing_converters.html">
Writing Converters
</a>
<ul class="md-nav__list">
<li class="md-nav__item">
<a class="md-nav__link" href="../contributors/writing_converters.html#background">
Background
</a>
</li>
<li class="md-nav__item">
<a class="md-nav__link" href="../contributors/writing_converters.html#converters">
Converters
</a>
</li>
<li class="md-nav__item">
<a class="md-nav__link" href="../contributors/writing_converters.html#converter-contract">
Converter Contract
</a>
</li>
<li class="md-nav__item">
<a class="md-nav__link" href="../contributors/writing_converters.html#conversion-context">
Conversion Context
</a>
</li>
<li class="md-nav__item">
<a class="md-nav__link" href="../contributors/writing_converters.html#args">
Args
</a>
</li>
<li class="md-nav__item">
<a class="md-nav__link" href="../contributors/writing_converters.html#weights">
Weights
</a>
</li>
<li class="md-nav__item">
<a class="md-nav__link" href="../contributors/writing_converters.html#other-advice">
Other advice
</a>
</li>
</ul>
</li>
<li class="md-nav__item">
<a class="md-nav__link" href="../contributors/useful_links.html">
Useful Links for TRTorch Development
</a>
<ul class="md-nav__list">
<li class="md-nav__item">
<a class="md-nav__link" href="../contributors/useful_links.html#tensorrt-available-layers-and-expected-dimensions">
TensorRT Available Layers and Expected Dimensions:
</a>
</li>
<li class="md-nav__item">
<a class="md-nav__link" href="../contributors/useful_links.html#tensorrt-c-documentation">
TensorRT C++ Documentation:
</a>
</li>
<li class="md-nav__item">
<a class="md-nav__link" href="../contributors/useful_links.html#tensorrt-python-documentation-sometimes-easier-to-read">
TensorRT Python Documentation (Sometimes easier to read):
</a>
</li>
<li class="md-nav__item">
<a class="md-nav__link" href="../contributors/useful_links.html#pytorch-functional-api">
PyTorch Functional API:
</a>
</li>
<li class="md-nav__item">
<a class="md-nav__link" href="../contributors/useful_links.html#pytorch-native-ops">
PyTorch native_ops:
</a>
</li>
<li class="md-nav__item">
<a class="md-nav__link" href="../contributors/useful_links.html#pytorch-ir-documentation">
PyTorch IR Documentation:
</a>
</li>
</ul>
</li>
<li class="md-nav__item">
<span class="md-nav__link caption">
<span class="caption-text">
Python API Documenation
</span>
</span>
</li>
<li class="md-nav__item">
<a class="md-nav__link" href="../py_api/trtorch.html">
trtorch
</a>
<ul class="md-nav__list">
<li class="md-nav__item">
<a class="md-nav__link" href="../py_api/trtorch.html#functions">
Functions
</a>
</li>
<li class="md-nav__item">
<a class="md-nav__link" href="../py_api/trtorch.html#enums">
Enums
</a>
</li>
<li class="md-nav__item">
<a class="md-nav__link" href="../py_api/trtorch.html#submodules">
Submodules
</a>
</li>
</ul>
</li>
<li class="md-nav__item">
<a class="md-nav__link" href="../py_api/logging.html">
trtorch.logging
</a>
</li>
<li class="md-nav__item">
<span class="md-nav__link caption">
<span class="caption-text">
C++ API Documenation
</span>
</span>
</li>
<li class="md-nav__item">
<a class="md-nav__link" href="../_cpp_api/trtorch_cpp.html">
TRTorch C++ API
</a>
<ul class="md-nav__list">
<li class="md-nav__item">
<a class="md-nav__link" href="../_cpp_api/trtorch_cpp.html#class-hierarchy">
Class Hierarchy
</a>
</li>
<li class="md-nav__item">
<a class="md-nav__link" href="../_cpp_api/trtorch_cpp.html#file-hierarchy">
File Hierarchy
</a>
</li>
<li class="md-nav__item">
<a class="md-nav__link" href="../_cpp_api/trtorch_cpp.html#full-api">
Full API
</a>
</li>
</ul>
</li>
</ul>
</nav>
</div>
</div>
</div>
<div class="md-sidebar md-sidebar--secondary" data-md-component="toc">
<div class="md-sidebar__scrollwrap">
<div class="md-sidebar__inner">
<nav class="md-nav md-nav--secondary">
<label class="md-nav__title" for="__toc">
Contents
</label>
<ul class="md-nav__list" data-md-scrollfix="">
<li class="md-nav__item">
<a class="md-nav__link" href="#tutorials-getting-started--page-root">
Getting Started
</a>
<nav class="md-nav">
<ul class="md-nav__list">
<li class="md-nav__item">
<a class="md-nav__link" href="#creating-a-torchscript-module">
Creating a TorchScript Module
</a>
</li>
<li class="md-nav__item">
<a class="md-nav__link" href="#working-with-torchscript-in-python">
Working with TorchScript in Python
</a>
</li>
<li class="md-nav__item">
<a class="md-nav__link" href="#compiling-with-trtorch-in-python">
Compiling with TRTorch in Python
</a>
</li>
<li class="md-nav__item">
<a class="md-nav__link" href="#working-with-torchscript-in-c">
Working with TorchScript in C++
</a>
</li>
<li class="md-nav__item">
<a class="md-nav__link" href="#compiling-with-trtorch-in-c">
Compiling with TRTorch in C++
</a>
</li>
<li class="md-nav__item">
<a class="md-nav__link" href="#under-the-hood">
Under The Hood
</a>
</li>
<li class="md-nav__item">
<a class="md-nav__link" href="#working-with-unsupported-operators">
Working with Unsupported Operators
</a>
<nav class="md-nav">
<ul class="md-nav__list">
<li class="md-nav__item">
<a class="md-nav__link" href="#registering-custom-converters">
Registering Custom Converters
</a>
</li>
</ul>
</nav>
</li>
</ul>
</nav>
</li>
<li class="md-nav__item">
<a class="md-nav__extra_link" href="../_sources/tutorials/getting_started.rst.txt">
Show Source
</a>
</li>
<li class="md-nav__item" id="searchbox">
</li>
</ul>
</nav>
</div>
</div>
</div>
<div class="md-content">
<article class="md-content__inner md-typeset" role="main">
<span id="id1">
</span>
<h1 id="tutorials-getting-started--page-root">
Getting Started
<a class="headerlink" href="#tutorials-getting-started--page-root" title="Permalink to this headline">
¶
</a>
</h1>
<p>
If you haven’t already, aquire a tarball of the library by following the instructions in
<a class="reference internal" href="installation.html#installation">
<span class="std std-ref">
Installation
</span>
</a>
</p>
<h2 id="creating-a-torchscript-module">
Creating a TorchScript Module
<a class="headerlink" href="#creating-a-torchscript-module" title="Permalink to this headline">
¶
</a>
</h2>
<p>
Once you have a trained model you want to compile with TRTorch, you need to start by converting that model from Python code to TorchScript code.
PyTorch has detailed documentation on how to do this
<a class="reference external" href="https://pytorch.org/tutorials/beginner/Intro_to_TorchScript_tutorial.html">
https://pytorch.org/tutorials/beginner/Intro_to_TorchScript_tutorial.html
</a>
but briefly here is the
here is key background and the process:
</p>
<p>
PyTorch programs are based around
<a href="#id2">
<span class="problematic" id="id3">
`
</span>
</a>
Module`s which can be used to compose higher level modules. Modules contain a constructor to set up the modules, parameters and sub-modules
and a forward function which describes how to use the parameters and submodules when the module is invoked.
</p>
<p>
For example, we can define a LeNet module like this:
</p>
<div class="highlight-python notranslate">
<table class="highlighttable">
<tr>
<td class="linenos">
<div class="linenodiv">
<pre> 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</pre>
</div>
</td>
<td class="code">
<div class="highlight">
<pre><span></span><span class="k">class</span> <span class="nc">LeNetFeatExtractor</span><span class="p">(</span><span class="n">nn</span><span class="o">.</span><span class="n">Module</span><span class="p">):</span>
<span class="k">def</span> <span class="fm">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span>
<span class="nb">super</span><span class="p">(</span><span class="n">LeNetFeatExtractor</span><span class="p">,</span> <span class="bp">self</span><span class="p">)</span><span class="o">.</span><span class="fm">__init__</span><span class="p">()</span>
<span class="bp">self</span><span class="o">.</span><span class="n">conv1</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Conv2d</span><span class="p">(</span><span class="mi">1</span><span class="p">,</span> <span class="mi">6</span><span class="p">,</span> <span class="mi">3</span><span class="p">)</span>
<span class="bp">self</span><span class="o">.</span><span class="n">conv2</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Conv2d</span><span class="p">(</span><span class="mi">6</span><span class="p">,</span> <span class="mi">16</span><span class="p">,</span> <span class="mi">3</span><span class="p">)</span>
<span class="k">def</span> <span class="nf">forward</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">x</span><span class="p">):</span>
<span class="n">x</span> <span class="o">=</span> <span class="n">F</span><span class="o">.</span><span class="n">max_pool2d</span><span class="p">(</span><span class="n">F</span><span class="o">.</span><span class="n">relu</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">conv1</span><span class="p">(</span><span class="n">x</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="n">x</span> <span class="o">=</span> <span class="n">F</span><span class="o">.</span><span class="n">max_pool2d</span><span class="p">(</span><span class="n">F</span><span class="o">.</span><span class="n">relu</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">conv2</span><span class="p">(</span><span class="n">x</span><span class="p">)),</span> <span class="mi">2</span><span class="p">)</span>
<span class="k">return</span> <span class="n">x</span>
<span class="k">class</span> <span class="nc">LeNetClassifier</span><span class="p">(</span><span class="n">nn</span><span class="o">.</span><span class="n">Module</span><span class="p">):</span>
<span class="k">def</span> <span class="fm">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span>
<span class="nb">super</span><span class="p">(</span><span class="n">LeNetClassifier</span><span class="p">,</span> <span class="bp">self</span><span class="p">)</span><span class="o">.</span><span class="fm">__init__</span><span class="p">()</span>
<span class="bp">self</span><span class="o">.</span><span class="n">fc1</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Linear</span><span class="p">(</span><span class="mi">16</span> <span class="o">*</span> <span class="mi">6</span> <span class="o">*</span> <span class="mi">6</span><span class="p">,</span> <span class="mi">120</span><span class="p">)</span>
<span class="bp">self</span><span class="o">.</span><span class="n">fc2</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Linear</span><span class="p">(</span><span class="mi">120</span><span class="p">,</span> <span class="mi">84</span><span class="p">)</span>
<span class="bp">self</span><span class="o">.</span><span class="n">fc3</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Linear</span><span class="p">(</span><span class="mi">84</span><span class="p">,</span> <span class="mi">10</span><span class="p">)</span>
<span class="k">def</span> <span class="nf">forward</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">x</span><span class="p">):</span>
<span class="n">x</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">flatten</span><span class="p">(</span><span class="n">x</span><span class="p">,</span><span class="mi">1</span><span class="p">)</span>
<span class="n">x</span> <span class="o">=</span> <span class="n">F</span><span class="o">.</span><span class="n">relu</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">fc1</span><span class="p">(</span><span class="n">x</span><span class="p">))</span>
<span class="n">x</span> <span class="o">=</span> <span class="n">F</span><span class="o">.</span><span class="n">relu</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">fc2</span><span class="p">(</span><span class="n">x</span><span class="p">))</span>
<span class="n">x</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">fc3</span><span class="p">(</span><span class="n">x</span><span class="p">)</span>
<span class="k">return</span> <span class="n">x</span>
<span class="k">class</span> <span class="nc">LeNet</span><span class="p">(</span><span class="n">nn</span><span class="o">.</span><span class="n">Module</span><span class="p">):</span>
<span class="k">def</span> <span class="fm">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span>
<span class="nb">super</span><span class="p">(</span><span class="n">LeNet</span><span class="p">,</span> <span class="bp">self</span><span class="p">)</span><span class="o">.</span><span class="fm">__init__</span><span class="p">()</span>
<span class="bp">self</span><span class="o">.</span><span class="n">feat</span> <span class="o">=</span> <span class="n">LeNetFeatExtractor</span><span class="p">()</span>
<span class="bp">self</span><span class="o">.</span><span class="n">classifer</span> <span class="o">=</span> <span class="n">LeNetClassifier</span><span class="p">()</span>
<span class="k">def</span> <span class="nf">forward</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">x</span><span class="p">):</span>
<span class="n">x</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">feat</span><span class="p">(</span><span class="n">x</span><span class="p">)</span>
<span class="n">x</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">classifer</span><span class="p">(</span><span class="n">x</span><span class="p">)</span>
<span class="k">return</span> <span class="n">x</span>
</pre>
</div>
</td>
</tr>
</table>
</div>
<p>
.
</p>
<blockquote>
<div>
<p>
Obviously you may want to consolidate such a simple model into a single module but we can see the composability of PyTorch here
</p>
</div>
</blockquote>
<p>
From here are two pathways for going from PyTorch Python code to TorchScript code: Tracing and Scripting.
</p>
<p>
Tracing follows the path of execution when the module is called and records what happens.
To trace an instance of our LeNet module, we can call
<code class="docutils literal notranslate">
<span class="pre">
torch.jit.trace
</span>
</code>
with an example input.
</p>
<div class="highlight-python notranslate">
<div class="highlight">
<pre><span></span><span class="n">model</span> <span class="o">=</span> <span class="n">LeNet</span><span class="p">()</span>
<span class="n">traced_model</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">jit</span><span class="o">.</span><span class="n">trace</span><span class="p">(</span><span class="n">model</span><span class="p">,</span> <span class="n">torch</span><span class="o">.</span><span class="n">empty</span><span class="p">([</span><span class="mi">1</span><span class="p">,</span><span class="mi">1</span><span class="p">,</span><span class="mi">32</span><span class="p">,</span><span class="mi">32</span><span class="p">]))</span>
</pre>
</div>
</div>
<p>
Scripting actually inspects your code with a compiler and generates an equivalent TorchScript program. The difference is that since tracing
is following the execution of your module, it cannot pick up control flow for instance. By working from the Python code, the compiler can
include these components. We can run the script compiler on our LeNet module by calling
<code class="docutils literal notranslate">
<span class="pre">
torch.jit.script
</span>
</code>
</p>
<div class="highlight-python notranslate">
<div class="highlight">
<pre><span></span><span class="n">model</span> <span class="o">=</span> <span class="n">LeNet</span><span class="p">()</span>
<span class="n">script_model</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">jit</span><span class="o">.</span><span class="n">script</span><span class="p">(</span><span class="n">model</span><span class="p">)</span>
</pre>
</div>
</div>
<p>
There are reasons to use one path or another, the PyTorch documentation has information on how to choose. From a TRTorch prespective, there is
better support (i.e your module is more likely to compile) for traced modules because it doesn’t include all the complexities of a complete
programming language, though both paths supported.
</p>
<p>
After scripting or tracing your module, you are given back a TorchScript Module. This contains the code and parameters used to run the module stored
in a intermediate representation that TRTorch can consume.
</p>
<p>
Here is what the LeNet traced module IR looks like:
</p>
<div class="highlight-none notranslate">
<div class="highlight">
<pre><span></span>graph(%self.1 : __torch__.___torch_mangle_10.LeNet,
%input.1 : Float(1, 1, 32, 32)):
%129 : __torch__.___torch_mangle_9.LeNetClassifier = prim::GetAttr[name="classifer"](%self.1)
%119 : __torch__.___torch_mangle_5.LeNetFeatExtractor = prim::GetAttr[name="feat"](%self.1)
%137 : Tensor = prim::CallMethod[name="forward"](%119, %input.1)
%138 : Tensor = prim::CallMethod[name="forward"](%129, %137)
return (%138)
</pre>
</div>
</div>
<p>
and the LeNet scripted module IR:
</p>
<div class="highlight-none notranslate">
<div class="highlight">
<pre><span></span>graph(%self : __torch__.LeNet,
%x.1 : Tensor):
%2 : __torch__.LeNetFeatExtractor = prim::GetAttr[name="feat"](%self)
%x.3 : Tensor = prim::CallMethod[name="forward"](%2, %x.1) # x.py:38:12
%5 : __torch__.LeNetClassifier = prim::GetAttr[name="classifer"](%self)
%x.5 : Tensor = prim::CallMethod[name="forward"](%5, %x.3) # x.py:39:12
return (%x.5)
</pre>
</div>
</div>
<p>
You can see that the IR preserves the module structure we have in our python code.
</p>
<span id="ts-in-py">
</span>
<h2 id="working-with-torchscript-in-python">
Working with TorchScript in Python
<a class="headerlink" href="#working-with-torchscript-in-python" title="Permalink to this headline">
¶
</a>
</h2>
<p>
TorchScript Modules are run the same way you run normal PyTorch modules. You can run the forward pass using the
<code class="docutils literal notranslate">
<span class="pre">
forward
</span>
</code>
method or just calling the module
<code class="docutils literal notranslate">
<span class="pre">
torch_scirpt_module(in_tensor)
</span>
</code>
The JIT compiler will compile
and optimize the module on the fly and then returns the results.
</p>
<span id="compile-py">
</span>
<h2 id="compiling-with-trtorch-in-python">
Compiling with TRTorch in Python
<a class="headerlink" href="#compiling-with-trtorch-in-python" title="Permalink to this headline">
¶
</a>
</h2>
<p>
To compile your TorchScript module with TRTorch, all you need to do is provide the module and some compiler settings
to TRTorch and you will be returned an optimized TorchScript module to run or add into another PyTorch module. The
only required setting is the input size or input range which is defined as a list of either list types like
<code class="docutils literal notranslate">
<span class="pre">
lists
</span>
</code>
,
<code class="docutils literal notranslate">
<span class="pre">
tuples
</span>
</code>
or PyTorch
<code class="docutils literal notranslate">
<span class="pre">
size
</span>
</code>
objects or dictionaries of minimum, optimial and maximum sizes. You can also specify settings such as
operating precision for the engine or target device.
</p>
<div class="highlight-python notranslate">
<div class="highlight">
<pre><span></span><span class="kn">import</span> <span class="nn">trtorch</span>
<span class="o">...</span>
<span class="n">compile_settings</span> <span class="o">=</span> <span class="p">{</span>
<span class="s2">"input_shapes"</span><span class="p">:</span> <span class="p">[</span>
<span class="p">{</span>
<span class="s2">"min"</span><span class="p">:</span> <span class="p">[</span><span class="mi">1</span><span class="p">,</span> <span class="mi">3</span><span class="p">,</span> <span class="mi">224</span><span class="p">,</span> <span class="mi">224</span><span class="p">],</span>
<span class="s2">"opt"</span><span class="p">:</span> <span class="p">[</span><span class="mi">1</span><span class="p">,</span> <span class="mi">3</span><span class="p">,</span> <span class="mi">512</span><span class="p">,</span> <span class="mi">512</span><span class="p">],</span>
<span class="s2">"max"</span><span class="p">:</span> <span class="p">[</span><span class="mi">1</span><span class="p">,</span> <span class="mi">3</span><span class="p">,</span> <span class="mi">1024</span><span class="p">,</span> <span class="mi">1024</span><span class="p">]</span>
<span class="p">},</span> <span class="c1"># For static size [1, 3, 224, 224]</span>
<span class="p">],</span>
<span class="s2">"op_precision"</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">half</span> <span class="c1"># Run with FP16</span>
<span class="p">}</span>
<span class="n">trt_ts_module</span> <span class="o">=</span> <span class="n">trtorch</span><span class="o">.</span><span class="n">compile</span><span class="p">(</span><span class="n">torch_script_module</span><span class="p">,</span> <span class="n">compile_settings</span><span class="p">)</span>
<span class="n">input_data</span> <span class="o">=</span> <span class="n">input_data</span><span class="o">.</span><span class="n">half</span><span class="p">()</span>
<span class="n">result</span> <span class="o">=</span> <span class="n">trt_ts_module</span><span class="p">(</span><span class="n">input_data</span><span class="p">)</span>
</pre>
</div>
</div>
<span id="ts-in-cc">
</span>
<h2 id="working-with-torchscript-in-c">
Working with TorchScript in C++
<a class="headerlink" href="#working-with-torchscript-in-c" title="Permalink to this headline">
¶
</a>
</h2>
<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="cpf"><torch/script.h> // One-stop header.</span><span class="cp"></span>
<span class="cp">#include</span> <span class="cpf"><iostream></span><span class="cp"></span>
<span class="cp">#include</span> <span class="cpf"><memory></span><span class="cp"></span>
<span class="kt">int</span> <span class="nf">main</span><span class="p">(</span><span class="kt">int</span> <span class="n">argc</span><span class="p">,</span> <span class="k">const</span> <span class="kt">char</span><span class="o">*</span> <span class="n">argv</span><span class="p">[])</span> <span class="p">{</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="n">module</span><span class="p">;</span>
<span class="k">try</span> <span class="p">{</span>
<span class="c1">// Deserialize the ScriptModule from a file using torch::jit::load().</span>
<span class="n">module</span> <span class="o">=</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="p">}</span>
<span class="k">catch</span> <span class="p">(</span><span class="k">const</span> <span class="n">c10</span><span class="o">::</span><span class="n">Error</span><span class="o">&</span> <span class="n">e</span><span class="p">)</span> <span class="p">{</span>
<span class="n">std</span><span class="o">::</span><span class="n">cerr</span> <span class="o"><<</span> <span class="s">"error loading the model</span><span class="se">\n</span><span class="s">"</span><span class="p">;</span>
<span class="k">return</span> <span class="o">-</span><span class="mi">1</span><span class="p">;</span>
<span class="p">}</span>
<span class="n">std</span><span class="o">::</span><span class="n">cout</span> <span class="o"><<</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="n">in</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">1</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="mi">32</span><span class="p">,</span> <span class="mi">32</span><span class="p">});</span>
<span class="k">auto</span> <span class="n">out</span> <span class="o">=</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="n">in</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">1</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="mi">32</span><span class="p">,</span> <span class="mi">32</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="k">auto</span> <span class="n">out</span> <span class="o">=</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>
<span id="compile-cpp">
</span>
<h2 id="compiling-with-trtorch-in-c">
Compiling with TRTorch in C++
<a class="headerlink" href="#compiling-with-trtorch-in-c" title="Permalink to this headline">
¶
</a>
</h2>
<p>