TensorRT 10.1.0
NvInfer.h
Go to the documentation of this file.
1/*
2 * SPDX-FileCopyrightText: Copyright (c) 1993-2024 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
3 * SPDX-License-Identifier: Apache-2.0
4 *
5 * Licensed under the Apache License, Version 2.0 (the "License");
6 * you may not use this file except in compliance with the License.
7 * You may obtain a copy of the License at
8 *
9 * http://www.apache.org/licenses/LICENSE-2.0
10 *
11 * Unless required by applicable law or agreed to in writing, software
12 * distributed under the License is distributed on an "AS IS" BASIS,
13 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
14 * See the License for the specific language governing permissions and
15 * limitations under the License.
16 */
17
18#ifndef NV_INFER_H
19#define NV_INFER_H
20
21#include "NvInferLegacyDims.h"
22#include "NvInferRuntime.h"
23
32//
35
41
47namespace nvinfer1
48{
49
57enum class LayerType : int32_t
58{
59 kCONVOLUTION = 0,
60 kCAST = 1,
61 kACTIVATION = 2,
62 kPOOLING = 3,
63 kLRN = 4,
64 kSCALE = 5,
65 kSOFTMAX = 6,
66 kDECONVOLUTION = 7,
67 kCONCATENATION = 8,
68 kELEMENTWISE = 9,
69 kPLUGIN = 10,
70 kUNARY = 11,
71 kPADDING = 12,
72 kSHUFFLE = 13,
73 kREDUCE = 14,
74 kTOPK = 15,
75 kGATHER = 16,
76 kMATRIX_MULTIPLY = 17,
77 kRAGGED_SOFTMAX = 18,
78 kCONSTANT = 19,
79 kIDENTITY = 20,
80 kPLUGIN_V2 = 21,
81 kSLICE = 22,
82 kSHAPE = 23,
83 kPARAMETRIC_RELU = 24,
84 kRESIZE = 25,
85 kTRIP_LIMIT = 26,
86 kRECURRENCE = 27,
87 kITERATOR = 28,
88 kLOOP_OUTPUT = 29,
89 kSELECT = 30,
90 kFILL = 31,
91 kQUANTIZE = 32,
92 kDEQUANTIZE = 33,
93 kCONDITION = 34,
96 kSCATTER = 37,
97 kEINSUM = 38,
98 kASSERTION = 39,
99 kONE_HOT = 40,
100 kNON_ZERO = 41,
101 kGRID_SAMPLE = 42,
102 kNMS = 43,
103 kREVERSE_SEQUENCE = 44,
104 kNORMALIZATION = 45,
105 kPLUGIN_V3 = 46
106};
107
113template <>
114constexpr inline int32_t EnumMax<LayerType>() noexcept
115{
116 return 47;
117}
118
125using TensorFormats = uint32_t;
126
132enum class ActivationType : int32_t
133{
134 kRELU = 0,
135 kSIGMOID = 1,
136 kTANH = 2,
137 kLEAKY_RELU = 3,
138 kELU = 4,
139 kSELU = 5,
140 kSOFTSIGN = 6,
141 kSOFTPLUS = 7,
142 kCLIP = 8,
143 kHARD_SIGMOID = 9,
144 kSCALED_TANH = 10,
145 kTHRESHOLDED_RELU = 11,
146 kGELU_ERF = 12,
147 kGELU_TANH = 13
148};
149
150namespace impl
151{
157template <>
159{
160 static constexpr int32_t kVALUE = 14;
161};
162} // namespace impl
163
180class ITensor : public INoCopy
181{
182public:
197 void setName(char const* name) noexcept
198 {
199 mImpl->setName(name);
200 }
201
209 char const* getName() const noexcept
210 {
211 return mImpl->getName();
212 }
213
228 void setDimensions(Dims const& dimensions) noexcept
229 {
230 mImpl->setDimensions(dimensions);
231 }
232
242 Dims getDimensions() const noexcept
243 {
244 return mImpl->getDimensions();
245 }
246
257 void setType(DataType type) noexcept
258 {
259 mImpl->setType(type);
260 }
261
269 DataType getType() const noexcept
270 {
271 return mImpl->getType();
272 }
273
286 TRT_DEPRECATED bool setDynamicRange(float min, float max) noexcept
287 {
288 return mImpl->setDynamicRange(min, max);
289 }
290
294 bool isNetworkInput() const noexcept
295 {
296 return mImpl->isNetworkInput();
297 }
298
302 bool isNetworkOutput() const noexcept
303 {
304 return mImpl->isNetworkOutput();
305 }
306
319 TRT_DEPRECATED void setBroadcastAcrossBatch(bool broadcastAcrossBatch) noexcept
320 {
321 mImpl->setBroadcastAcrossBatch(broadcastAcrossBatch);
322 }
323
334 {
335 return mImpl->getBroadcastAcrossBatch();
336 }
337
346 {
347 return mImpl->getLocation();
348 }
349
365 {
366 mImpl->setLocation(location);
367 }
368
376 bool dynamicRangeIsSet() const noexcept
377 {
378 return mImpl->dynamicRangeIsSet();
379 }
380
384 void resetDynamicRange() noexcept
385 {
386 mImpl->resetDynamicRange();
387 }
388
394 float getDynamicRangeMin() const noexcept
395 {
396 return mImpl->getDynamicRangeMin();
397 }
398
404 float getDynamicRangeMax() const noexcept
405 {
406 return mImpl->getDynamicRangeMax();
407 }
408
426 void setAllowedFormats(TensorFormats formats) noexcept
427 {
428 mImpl->setAllowedFormats(formats);
429 }
430
440 {
441 return mImpl->getAllowedFormats();
442 }
443
470 bool isShapeTensor() const noexcept
471 {
472 return mImpl->isShapeTensor();
473 }
474
491 bool isExecutionTensor() const noexcept
492 {
493 return mImpl->isExecutionTensor();
494 }
495
517 void setDimensionName(int32_t index, char const* name) noexcept
518 {
519 mImpl->setDimensionName(index, name);
520 }
521
532 char const* getDimensionName(int32_t index) const noexcept
533 {
534 return mImpl->getDimensionName(index);
535 }
536
537protected:
538 apiv::VTensor* mImpl;
539 virtual ~ITensor() noexcept = default;
540};
541
549class ILayer : public INoCopy
550{
551public:
557 LayerType getType() const noexcept
558 {
559 return mLayer->getType();
560 }
561
571 void setName(char const* name) noexcept
572 {
573 mLayer->setName(name);
574 }
575
581 char const* getName() const noexcept
582 {
583 return mLayer->getName();
584 }
585
589 int32_t getNbInputs() const noexcept
590 {
591 return mLayer->getNbInputs();
592 }
593
602 ITensor* getInput(int32_t index) const noexcept
603 {
604 return mLayer->getInput(index);
605 }
606
610 int32_t getNbOutputs() const noexcept
611 {
612 return mLayer->getNbOutputs();
613 }
614
620 ITensor* getOutput(int32_t index) const noexcept
621 {
622 return mLayer->getOutput(index);
623 }
624
637 void setInput(int32_t index, ITensor& tensor) noexcept
638 {
639 return mLayer->setInput(index, tensor);
640 }
641
668 void setPrecision(DataType dataType) noexcept
669 {
670 mLayer->setPrecision(dataType);
671 }
672
680 DataType getPrecision() const noexcept
681 {
682 return mLayer->getPrecision();
683 }
684
692 bool precisionIsSet() const noexcept
693 {
694 return mLayer->precisionIsSet();
695 }
696
702 void resetPrecision() noexcept
703 {
704 mLayer->resetPrecision();
705 }
706
749 void setOutputType(int32_t index, DataType dataType) noexcept
750 {
751 mLayer->setOutputType(index, dataType);
752 }
753
764 DataType getOutputType(int32_t index) const noexcept
765 {
766 return mLayer->getOutputType(index);
767 }
768
778 bool outputTypeIsSet(int32_t index) const noexcept
779 {
780 return mLayer->outputTypeIsSet(index);
781 }
782
790 void resetOutputType(int32_t index) noexcept
791 {
792 return mLayer->resetOutputType(index);
793 }
794
808 void setMetadata(char const* metadata) noexcept
809 {
810 mLayer->setMetadata(metadata);
811 }
812
821 char const* getMetadata() const noexcept
822 {
823 return mLayer->getMetadata();
824 }
825
826protected:
827 virtual ~ILayer() noexcept = default;
828 apiv::VLayer* mLayer;
829};
830
987enum class PaddingMode : int32_t
988{
991 kSAME_UPPER = 2,
992 kSAME_LOWER = 3,
993};
994
995namespace impl
996{
1002template <>
1004{
1005 static constexpr int32_t kVALUE = 4;
1006};
1007} // namespace impl
1008
1022{
1023public:
1031 void setNbOutputMaps(int64_t nbOutputMaps) noexcept
1032 {
1033 mImpl->setNbOutputMaps(nbOutputMaps);
1034 }
1035
1041 int64_t getNbOutputMaps() const noexcept
1042 {
1043 return mImpl->getNbOutputMaps();
1044 }
1045
1061 void setNbGroups(int64_t nbGroups) noexcept
1062 {
1063 mImpl->setNbGroups(nbGroups);
1064 }
1065
1071 int64_t getNbGroups() const noexcept
1072 {
1073 return mImpl->getNbGroups();
1074 }
1075
1085 void setKernelWeights(Weights weights) noexcept
1086 {
1087 mImpl->setKernelWeights(weights);
1088 }
1089
1095 Weights getKernelWeights() const noexcept
1096 {
1097 return mImpl->getKernelWeights();
1098 }
1099
1110 void setBiasWeights(Weights weights) noexcept
1111 {
1112 mImpl->setBiasWeights(weights);
1113 }
1114
1120 Weights getBiasWeights() const noexcept
1121 {
1122 return mImpl->getBiasWeights();
1123 }
1124
1137 void setPrePadding(Dims const& padding) noexcept
1138 {
1139 mImpl->setPrePadding(padding);
1140 }
1141
1147 Dims getPrePadding() const noexcept
1148 {
1149 return mImpl->getPrePadding();
1150 }
1151
1164 void setPostPadding(Dims const& padding) noexcept
1165 {
1166 mImpl->setPostPadding(padding);
1167 }
1168
1174 Dims getPostPadding() const noexcept
1175 {
1176 return mImpl->getPostPadding();
1177 }
1178
1188 void setPaddingMode(PaddingMode paddingMode) noexcept
1189 {
1190 mImpl->setPaddingMode(paddingMode);
1191 }
1192
1201 {
1202 return mImpl->getPaddingMode();
1203 }
1204
1213 void setKernelSizeNd(Dims const& kernelSize) noexcept
1214 {
1215 mImpl->setKernelSizeNd(kernelSize);
1216 }
1217
1223 Dims getKernelSizeNd() const noexcept
1224 {
1225 return mImpl->getKernelSizeNd();
1226 }
1227
1238 void setStrideNd(Dims const& stride) noexcept
1239 {
1240 mImpl->setStrideNd(stride);
1241 }
1242
1248 Dims getStrideNd() const noexcept
1249 {
1250 return mImpl->getStrideNd();
1251 }
1252
1266 void setPaddingNd(Dims const& padding) noexcept
1267 {
1268 mImpl->setPaddingNd(padding);
1269 }
1270
1278 Dims getPaddingNd() const noexcept
1279 {
1280 return mImpl->getPaddingNd();
1281 }
1282
1292 void setDilationNd(Dims const& dilation) noexcept
1293 {
1294 mImpl->setDilationNd(dilation);
1295 }
1296
1302 Dims getDilationNd() const noexcept
1303 {
1304 return mImpl->getDilationNd();
1305 }
1306
1321 using ILayer::setInput;
1322
1323protected:
1324 virtual ~IConvolutionLayer() noexcept = default;
1325 apiv::VConvolutionLayer* mImpl;
1326};
1327
1342{
1343public:
1352 {
1353 mImpl->setActivationType(type);
1354 }
1355
1362 {
1363 return mImpl->getActivationType();
1364 }
1365
1376 void setAlpha(float alpha) noexcept
1377 {
1378 mImpl->setAlpha(alpha);
1379 }
1380
1390 void setBeta(float beta) noexcept
1391 {
1392 mImpl->setBeta(beta);
1393 }
1394
1399 float getAlpha() const noexcept
1400 {
1401 return mImpl->getAlpha();
1402 }
1403
1408 float getBeta() const noexcept
1409 {
1410 return mImpl->getBeta();
1411 }
1412
1413protected:
1414 virtual ~IActivationLayer() noexcept = default;
1415 apiv::VActivationLayer* mImpl;
1416};
1417
1423enum class PoolingType : int32_t
1424{
1425 kMAX = 0,
1426 kAVERAGE = 1,
1428};
1429
1430namespace impl
1431{
1437template <>
1439{
1440 static constexpr int32_t kVALUE = 3;
1441};
1442} // namespace impl
1443
1455class IPoolingLayer : public ILayer
1456{
1457public:
1465 void setPoolingType(PoolingType type) noexcept
1466 {
1467 mImpl->setPoolingType(type);
1468 }
1469
1476 {
1477 return mImpl->getPoolingType();
1478 }
1479
1490 void setBlendFactor(float blendFactor) noexcept
1491 {
1492 mImpl->setBlendFactor(blendFactor);
1493 }
1494
1503 float getBlendFactor() const noexcept
1504 {
1505 return mImpl->getBlendFactor();
1506 }
1507
1517 void setAverageCountExcludesPadding(bool exclusive) noexcept
1518 {
1519 mImpl->setAverageCountExcludesPadding(exclusive);
1520 }
1521
1529 {
1530 return mImpl->getAverageCountExcludesPadding();
1531 }
1532
1546 void setPrePadding(Dims const& padding) noexcept
1547 {
1548 mImpl->setPrePadding(padding);
1549 }
1550
1556 Dims getPrePadding() const noexcept
1557 {
1558 return mImpl->getPrePadding();
1559 }
1560
1574 void setPostPadding(Dims const& padding) noexcept
1575 {
1576 mImpl->setPostPadding(padding);
1577 }
1578
1584 Dims getPostPadding() const noexcept
1585 {
1586 return mImpl->getPostPadding();
1587 }
1588
1597 void setPaddingMode(PaddingMode paddingMode) noexcept
1598 {
1599 mImpl->setPaddingMode(paddingMode);
1600 }
1601
1609 {
1610 return mImpl->getPaddingMode();
1611 }
1612
1621 void setWindowSizeNd(Dims const& windowSize) noexcept
1622 {
1623 mImpl->setWindowSizeNd(windowSize);
1624 }
1625
1631 Dims getWindowSizeNd() const noexcept
1632 {
1633 return mImpl->getWindowSizeNd();
1634 }
1635
1646 void setStrideNd(Dims const& stride) noexcept
1647 {
1648 mImpl->setStrideNd(stride);
1649 }
1650
1656 Dims getStrideNd() const noexcept
1657 {
1658 return mImpl->getStrideNd();
1659 }
1660
1675 void setPaddingNd(Dims const& padding) noexcept
1676 {
1677 mImpl->setPaddingNd(padding);
1678 }
1679
1687 Dims getPaddingNd() const noexcept
1688 {
1689 return mImpl->getPaddingNd();
1690 }
1691
1692protected:
1693 virtual ~IPoolingLayer() noexcept = default;
1694 apiv::VPoolingLayer* mImpl;
1695};
1696
1706class ILRNLayer : public ILayer
1707{
1708public:
1718 void setWindowSize(int64_t windowSize) noexcept
1719 {
1720 mImpl->setWindowSize(windowSize);
1721 }
1722
1728 int64_t getWindowSize() const noexcept
1729 {
1730 return mImpl->getWindowSize();
1731 }
1732
1740 void setAlpha(float alpha) noexcept
1741 {
1742 mImpl->setAlpha(alpha);
1743 }
1744
1750 float getAlpha() const noexcept
1751 {
1752 return mImpl->getAlpha();
1753 }
1754
1762 void setBeta(float beta) noexcept
1763 {
1764 mImpl->setBeta(beta);
1765 }
1766
1772 float getBeta() const noexcept
1773 {
1774 return mImpl->getBeta();
1775 }
1776
1784 void setK(float k) noexcept
1785 {
1786 mImpl->setK(k);
1787 }
1788
1794 float getK() const noexcept
1795 {
1796 return mImpl->getK();
1797 }
1798
1799protected:
1800 virtual ~ILRNLayer() noexcept = default;
1801 apiv::VLRNLayer* mImpl;
1802};
1803
1809enum class ScaleMode : int32_t
1810{
1811 kUNIFORM = 0,
1812 kCHANNEL = 1,
1813 kELEMENTWISE = 2
1814};
1815
1821template <>
1822constexpr inline int32_t EnumMax<ScaleMode>() noexcept
1823{
1824 return 3;
1825}
1826
1852class IScaleLayer : public ILayer
1853{
1854public:
1860 void setMode(ScaleMode mode) noexcept
1861 {
1862 mImpl->setMode(mode);
1863 }
1864
1870 ScaleMode getMode() const noexcept
1871 {
1872 return mImpl->getMode();
1873 }
1874
1880 void setShift(Weights shift) noexcept
1881 {
1882 mImpl->setShift(shift);
1883 }
1884
1890 Weights getShift() const noexcept
1891 {
1892 return mImpl->getShift();
1893 }
1894
1900 void setScale(Weights scale) noexcept
1901 {
1902 mImpl->setScale(scale);
1903 }
1904
1910 Weights getScale() const noexcept
1911 {
1912 return mImpl->getScale();
1913 }
1914
1920 void setPower(Weights power) noexcept
1921 {
1922 mImpl->setPower(power);
1923 }
1924
1930 Weights getPower() const noexcept
1931 {
1932 return mImpl->getPower();
1933 }
1934
1945 int32_t getChannelAxis() const noexcept
1946 {
1947 return mImpl->getChannelAxis();
1948 }
1949
1966 void setChannelAxis(int32_t channelAxis) noexcept
1967 {
1968 mImpl->setChannelAxis(channelAxis);
1969 }
1970
1971protected:
1972 virtual ~IScaleLayer() noexcept = default;
1973 apiv::VScaleLayer* mImpl;
1974};
1975
1996class ISoftMaxLayer : public ILayer
1997{
1998public:
2019 void setAxes(uint32_t axes) noexcept
2020 {
2021 mImpl->setAxes(axes);
2022 }
2023
2029 uint32_t getAxes() const noexcept
2030 {
2031 return mImpl->getAxes();
2032 }
2033
2034protected:
2035 virtual ~ISoftMaxLayer() noexcept = default;
2036 apiv::VSoftMaxLayer* mImpl;
2037};
2038
2052{
2053public:
2065 void setAxis(int32_t axis) noexcept
2066 {
2067 mImpl->setAxis(axis);
2068 }
2069
2075 int32_t getAxis() const noexcept
2076 {
2077 return mImpl->getAxis();
2078 }
2079
2080protected:
2081 virtual ~IConcatenationLayer() noexcept = default;
2082 apiv::VConcatenationLayer* mImpl;
2083};
2084
2093{
2094public:
2102 void setNbOutputMaps(int64_t nbOutputMaps) noexcept
2103 {
2104 mImpl->setNbOutputMaps(nbOutputMaps);
2105 }
2106
2112 int64_t getNbOutputMaps() const noexcept
2113 {
2114 return mImpl->getNbOutputMaps();
2115 }
2116
2132 void setNbGroups(int64_t nbGroups) noexcept
2133 {
2134 mImpl->setNbGroups(nbGroups);
2135 }
2136
2142 int64_t getNbGroups() const noexcept
2143 {
2144 return mImpl->getNbGroups();
2145 }
2146
2156 void setKernelWeights(Weights weights) noexcept
2157 {
2158 mImpl->setKernelWeights(weights);
2159 }
2160
2166 Weights getKernelWeights() const noexcept
2167 {
2168 return mImpl->getKernelWeights();
2169 }
2170
2181 void setBiasWeights(Weights weights) noexcept
2182 {
2183 mImpl->setBiasWeights(weights);
2184 }
2185
2191 Weights getBiasWeights() const noexcept
2192 {
2193 return mImpl->getBiasWeights();
2194 }
2195
2209 void setPrePadding(Dims const& padding) noexcept
2210 {
2211 mImpl->setPrePadding(padding);
2212 }
2213
2219 Dims getPrePadding() const noexcept
2220 {
2221 return mImpl->getPrePadding();
2222 }
2223
2237 void setPostPadding(Dims const& padding) noexcept
2238 {
2239 mImpl->setPostPadding(padding);
2240 }
2241
2247 Dims getPostPadding() const noexcept
2248 {
2249 return mImpl->getPostPadding();
2250 }
2251
2261 void setPaddingMode(PaddingMode paddingMode) noexcept
2262 {
2263 mImpl->setPaddingMode(paddingMode);
2264 }
2265
2274 {
2275 return mImpl->getPaddingMode();
2276 }
2277
2288 void setKernelSizeNd(Dims const& kernelSize) noexcept
2289 {
2290 mImpl->setKernelSizeNd(kernelSize);
2291 }
2292
2298 Dims getKernelSizeNd() const noexcept
2299 {
2300 return mImpl->getKernelSizeNd();
2301 }
2302
2315 void setStrideNd(Dims const& stride) noexcept
2316 {
2317 mImpl->setStrideNd(stride);
2318 }
2319
2325 Dims getStrideNd() const noexcept
2326 {
2327 return mImpl->getStrideNd();
2328 }
2329
2343 void setPaddingNd(Dims const& padding) noexcept
2344 {
2345 mImpl->setPaddingNd(padding);
2346 }
2347
2355 Dims getPaddingNd() const noexcept
2356 {
2357 return mImpl->getPaddingNd();
2358 }
2359
2372 using ILayer::setInput;
2373
2381 void setDilationNd(Dims const& dilation) noexcept
2382 {
2383 mImpl->setDilationNd(dilation);
2384 }
2385
2391 Dims getDilationNd() const noexcept
2392 {
2393 return mImpl->getDilationNd();
2394 }
2395
2396protected:
2397 virtual ~IDeconvolutionLayer() noexcept = default;
2398 apiv::VDeconvolutionLayer* mImpl;
2399};
2400
2415enum class ElementWiseOperation : int32_t
2416{
2417 kSUM = 0,
2418 kPROD = 1,
2419 kMAX = 2,
2420 kMIN = 3,
2421 kSUB = 4,
2422 kDIV = 5,
2423 kPOW = 6,
2424 kFLOOR_DIV = 7,
2425 kAND = 8,
2426 kOR = 9,
2427 kXOR = 10,
2428 kEQUAL = 11,
2429 kGREATER = 12,
2430 kLESS = 13
2431};
2432
2433namespace impl
2434{
2440template <>
2442{
2443 static constexpr int32_t kVALUE = 14;
2444};
2445} // namespace impl
2446
2467{
2468public:
2479 {
2480 return mImpl->setOperation(op);
2481 }
2482
2491 {
2492 return mImpl->getOperation();
2493 }
2494
2495protected:
2496 apiv::VElementWiseLayer* mImpl;
2497 virtual ~IElementWiseLayer() noexcept = default;
2498};
2499
2505enum class GatherMode : int32_t
2506{
2507 kDEFAULT = 0,
2508 kELEMENT = 1,
2509 kND = 2
2510};
2511
2517template <>
2518constexpr inline int32_t EnumMax<GatherMode>() noexcept
2519{
2520 return 3;
2521}
2522
2600class IGatherLayer : public ILayer
2601{
2602public:
2612 void setGatherAxis(int32_t axis) noexcept
2613 {
2614 mImpl->setGatherAxis(axis);
2615 }
2616
2624 int32_t getGatherAxis() const noexcept
2625 {
2626 return mImpl->getGatherAxis();
2627 }
2628
2647 void setNbElementWiseDims(int32_t elementWiseDims) noexcept
2648 {
2649 mImpl->setNbElementWiseDims(elementWiseDims);
2650 }
2651
2657 int32_t getNbElementWiseDims() const noexcept
2658 {
2659 return mImpl->getNbElementWiseDims();
2660 }
2661
2667 void setMode(GatherMode mode) noexcept
2668 {
2669 mImpl->setMode(mode);
2670 }
2671
2677 GatherMode getMode() const noexcept
2678 {
2679 return mImpl->getMode();
2680 }
2681
2682protected:
2683 apiv::VGatherLayer* mImpl;
2684 virtual ~IGatherLayer() noexcept = default;
2685};
2686
2697{
2698public:
2705 {
2706 return mImpl->getPlugin();
2707 }
2708
2709protected:
2710 apiv::VPluginV2Layer* mImpl;
2711 virtual ~IPluginV2Layer() noexcept = default;
2712};
2713
2724{
2725public:
2732 {
2733 return mImpl->getPlugin();
2734 }
2735
2736protected:
2737 apiv::VPluginV3Layer* mImpl;
2738 virtual ~IPluginV3Layer() noexcept = default;
2739};
2740
2757enum class UnaryOperation : int32_t
2758{
2759 kEXP = 0,
2760 kLOG = 1,
2761 kSQRT = 2,
2762 kRECIP = 3,
2763 kABS = 4,
2764 kNEG = 5,
2765 kSIN = 6,
2766 kCOS = 7,
2767 kTAN = 8,
2768 kSINH = 9,
2769 kCOSH = 10,
2770 kASIN = 11,
2771 kACOS = 12,
2772 kATAN = 13,
2773 kASINH = 14,
2774 kACOSH = 15,
2775 kATANH = 16,
2776 kCEIL = 17,
2777 kFLOOR = 18,
2778 kERF = 19,
2779 kNOT = 20,
2780 kSIGN = 21,
2781 kROUND = 22,
2782 kISINF = 23,
2783 kISNAN = 24,
2784};
2785
2791template <>
2792constexpr inline int32_t EnumMax<UnaryOperation>() noexcept
2793{
2794 return 25;
2795}
2796
2804class IUnaryLayer : public ILayer
2805{
2806public:
2815 {
2816 mImpl->setOperation(op);
2817 }
2818
2825 {
2826 return mImpl->getOperation();
2827 }
2828
2829protected:
2830 apiv::VUnaryLayer* mImpl;
2831 virtual ~IUnaryLayer() noexcept = default;
2832};
2833
2852enum class ReduceOperation : int32_t
2853{
2854 kSUM = 0,
2855 kPROD = 1,
2856 kMAX = 2,
2857 kMIN = 3,
2858 kAVG = 4
2859};
2860
2866template <>
2867constexpr inline int32_t EnumMax<ReduceOperation>() noexcept
2868{
2869 return 5;
2870}
2871
2879class IReduceLayer : public ILayer
2880{
2881public:
2888 {
2889 mImpl->setOperation(op);
2890 }
2891
2898 {
2899 return mImpl->getOperation();
2900 }
2901
2907 void setReduceAxes(uint32_t reduceAxes) noexcept
2908 {
2909 mImpl->setReduceAxes(reduceAxes);
2910 }
2911
2917 uint32_t getReduceAxes() const noexcept
2918 {
2919 return mImpl->getReduceAxes();
2920 }
2921
2927 void setKeepDimensions(bool keepDimensions) noexcept
2928 {
2929 mImpl->setKeepDimensions(keepDimensions);
2930 }
2931
2937 bool getKeepDimensions() const noexcept
2938 {
2939 return mImpl->getKeepDimensions();
2940 }
2941
2942protected:
2943 apiv::VReduceLayer* mImpl;
2944 virtual ~IReduceLayer() noexcept = default;
2945};
2946
2957class IPaddingLayer : public ILayer
2958{
2959public:
2969 void setPrePaddingNd(Dims const& padding) noexcept
2970 {
2971 mImpl->setPrePaddingNd(padding);
2972 }
2973
2981 Dims getPrePaddingNd() const noexcept
2982 {
2983 return mImpl->getPrePaddingNd();
2984 }
2985
2995 void setPostPaddingNd(Dims const& padding) noexcept
2996 {
2997 mImpl->setPostPaddingNd(padding);
2998 }
2999
3007 Dims getPostPaddingNd() const noexcept
3008 {
3009 return mImpl->getPostPaddingNd();
3010 }
3011
3012protected:
3013 apiv::VPaddingLayer* mImpl;
3014 virtual ~IPaddingLayer() noexcept = default;
3015};
3016
3023{
3030 int32_t order[Dims::MAX_DIMS];
3031};
3032
3045class IShuffleLayer : public ILayer
3046{
3047public:
3057 void setFirstTranspose(Permutation permutation) noexcept
3058 {
3059 mImpl->setFirstTranspose(permutation);
3060 }
3061
3070 {
3071 return mImpl->getFirstTranspose();
3072 }
3073
3097 void setReshapeDimensions(Dims const& dimensions) noexcept
3098 {
3099 mImpl->setReshapeDimensions(dimensions);
3100 }
3101
3111 {
3112 return mImpl->getReshapeDimensions();
3113 }
3114
3120 //
3143 using ILayer::setInput;
3144
3157 void setSecondTranspose(Permutation permutation) noexcept
3158 {
3159 mImpl->setSecondTranspose(permutation);
3160 }
3161
3170 {
3171 return mImpl->getSecondTranspose();
3172 }
3173
3185 void setZeroIsPlaceholder(bool zeroIsPlaceholder) noexcept
3186 {
3187 return mImpl->setZeroIsPlaceholder(zeroIsPlaceholder);
3188 }
3189
3198 bool getZeroIsPlaceholder() const noexcept
3199 {
3200 return mImpl->getZeroIsPlaceholder();
3201 }
3202
3203protected:
3204 apiv::VShuffleLayer* mImpl;
3205 virtual ~IShuffleLayer() noexcept = default;
3206};
3207
3213enum class SampleMode : int32_t
3214{
3215 kSTRICT_BOUNDS = 0,
3216 kWRAP = 1,
3217 kCLAMP = 2,
3218 kFILL = 3,
3219 kREFLECT = 4,
3222};
3223
3229template <>
3230constexpr inline int32_t EnumMax<SampleMode>() noexcept
3231{
3232 return 5;
3233}
3234
3277class ISliceLayer : public ILayer
3278{
3279public:
3289 void setStart(Dims const& start) noexcept
3290 {
3291 mImpl->setStart(start);
3292 }
3293
3304 Dims getStart() const noexcept
3305 {
3306 return mImpl->getStart();
3307 }
3308
3318 void setSize(Dims const& size) noexcept
3319 {
3320 return mImpl->setSize(size);
3321 }
3322
3333 Dims getSize() const noexcept
3334 {
3335 return mImpl->getSize();
3336 }
3337
3347 void setStride(Dims const& stride) noexcept
3348 {
3349 mImpl->setStride(stride);
3350 }
3351
3362 Dims getStride() const noexcept
3363 {
3364 return mImpl->getStride();
3365 }
3366
3372 void setMode(SampleMode mode) noexcept
3373 {
3374 mImpl->setMode(mode);
3375 }
3376
3382 SampleMode getMode() const noexcept
3383 {
3384 return mImpl->getMode();
3385 }
3386
3410 using ILayer::setInput;
3411
3412protected:
3413 apiv::VSliceLayer* mImpl;
3414 virtual ~ISliceLayer() noexcept = default;
3415};
3416
3429class IShapeLayer : public ILayer
3430{
3431protected:
3432 apiv::VShapeLayer* mImpl;
3433 virtual ~IShapeLayer() noexcept = default;
3434};
3435
3441enum class TopKOperation : int32_t
3442{
3443 kMAX = 0,
3444 kMIN = 1,
3445};
3446
3452template <>
3453constexpr inline int32_t EnumMax<TopKOperation>() noexcept
3454{
3455 return 2;
3456}
3457
3469class ITopKLayer : public ILayer
3470{
3471public:
3477 void setOperation(TopKOperation op) noexcept
3478 {
3479 mImpl->setOperation(op);
3480 }
3481
3488 {
3489 return mImpl->getOperation();
3490 }
3491
3501 void setK(int32_t k) noexcept
3502 {
3503 mImpl->setK(k);
3504 }
3505
3515 int32_t getK() const noexcept
3516 {
3517 return mImpl->getK();
3518 }
3519
3525 void setReduceAxes(uint32_t reduceAxes) noexcept
3526 {
3527 mImpl->setReduceAxes(reduceAxes);
3528 }
3529
3535 uint32_t getReduceAxes() const noexcept
3536 {
3537 return mImpl->getReduceAxes();
3538 }
3539
3554 using ILayer::setInput;
3555
3556protected:
3557 apiv::VTopKLayer* mImpl;
3558 virtual ~ITopKLayer() noexcept = default;
3559};
3560
3567enum class MatrixOperation : int32_t
3568{
3572 kNONE = 0,
3573
3575 kTRANSPOSE = 1,
3576
3587 kVECTOR = 2,
3588};
3589
3595template <>
3596constexpr inline int32_t EnumMax<MatrixOperation>() noexcept
3597{
3598 return 3;
3599}
3600
3627{
3628public:
3637 void setOperation(int32_t index, MatrixOperation op) noexcept
3638 {
3639 mImpl->setOperation(index, op);
3640 }
3641
3649 MatrixOperation getOperation(int32_t index) const noexcept
3650 {
3651 return mImpl->getOperation(index);
3652 }
3653
3654protected:
3655 apiv::VMatrixMultiplyLayer* mImpl;
3656 virtual ~IMatrixMultiplyLayer() noexcept = default;
3657};
3658
3680class INonZeroLayer : public ILayer
3681{
3682protected:
3683 virtual ~INonZeroLayer() noexcept = default;
3684 apiv::VNonZeroLayer* mImpl;
3685};
3686
3702{
3703protected:
3704 apiv::VRaggedSoftMaxLayer* mImpl;
3705 virtual ~IRaggedSoftMaxLayer() noexcept = default;
3706};
3707
3735{
3736protected:
3737 apiv::VIdentityLayer* mImpl;
3738 virtual ~IIdentityLayer() noexcept = default;
3739};
3740
3747class ICastLayer : public ILayer
3748{
3749public:
3757 void setToType(DataType toType) noexcept
3758 {
3759 mImpl->setToType(toType);
3760 }
3761
3768 DataType getToType() const noexcept
3769 {
3770 return mImpl->getToType();
3771 }
3772
3773protected:
3774 apiv::VCastLayer* mImpl;
3775 virtual ~ICastLayer() noexcept = default;
3776};
3777
3787{
3788public:
3797 void setWeights(Weights weights) noexcept
3798 {
3799 mImpl->setWeights(weights);
3800 }
3801
3807 Weights getWeights() const noexcept
3808 {
3809 return mImpl->getWeights();
3810 }
3811
3819 void setDimensions(Dims const& dimensions) noexcept
3820 {
3821 mImpl->setDimensions(dimensions);
3822 }
3823
3831 Dims getDimensions() const noexcept
3832 {
3833 return mImpl->getDimensions();
3834 }
3835
3836protected:
3837 apiv::VConstantLayer* mImpl;
3838 virtual ~IConstantLayer() noexcept = default;
3839};
3840
3851{
3852protected:
3853 apiv::VParametricReLULayer* mImpl;
3854 virtual ~IParametricReLULayer() noexcept = default;
3855};
3856
3862enum class InterpolationMode : int32_t
3863{
3864 kNEAREST = 0,
3865 kLINEAR = 1,
3866 kCUBIC = 2
3867};
3868
3869namespace impl
3870{
3876template <>
3878{
3879 static constexpr int32_t kVALUE = 3;
3880};
3881} // namespace impl
3882
3891{
3904 kALIGN_CORNERS = 0,
3905
3912 kASYMMETRIC = 1,
3913
3920 kHALF_PIXEL = 2,
3921};
3922
3923namespace impl
3924{
3930template <>
3932{
3933 static constexpr int32_t kVALUE = 3;
3934};
3935} // namespace impl
3936
3944enum class ResizeSelector : int32_t
3945{
3947 kFORMULA = 0,
3948
3950 kUPPER = 1,
3951};
3952
3953namespace impl
3954{
3960template <>
3962{
3963 static constexpr int32_t kVALUE = 2;
3964};
3965} // namespace impl
3966
3974enum class ResizeRoundMode : int32_t
3975{
3977 kHALF_UP = 0,
3978
3980 kHALF_DOWN = 1,
3981
3983 kFLOOR = 2,
3984
3986 kCEIL = 3,
3987};
3988
3989namespace impl
3990{
3996template <>
3998{
3999 static constexpr int32_t kVALUE = 4;
4000};
4001} // namespace impl
4002
4039class IResizeLayer : public ILayer
4040{
4041public:
4060 void setOutputDimensions(Dims const& dimensions) noexcept
4061 {
4062 return mImpl->setOutputDimensions(dimensions);
4063 }
4064
4070 Dims getOutputDimensions() const noexcept
4071 {
4072 return mImpl->getOutputDimensions();
4073 }
4074
4100 void setScales(float const* scales, int32_t nbScales) noexcept
4101 {
4102 mImpl->setScales(scales, nbScales);
4103 }
4104
4119 int32_t getScales(int32_t size, float* scales) const noexcept
4120 {
4121 return mImpl->getScales(size, scales);
4122 }
4123
4131 void setResizeMode(InterpolationMode interpolationMode) noexcept
4132 {
4133 mImpl->setResizeMode(interpolationMode);
4134 }
4135
4142 {
4143 return mImpl->getResizeMode();
4144 }
4145
4165 using ILayer::setInput;
4166
4177 {
4178 mImpl->setCoordinateTransformation(coordTransform);
4179 }
4180
4187 {
4188 return mImpl->getCoordinateTransformation();
4189 }
4190
4202 {
4203 mImpl->setSelectorForSinglePixel(selector);
4204 }
4205
4212 {
4213 return mImpl->getSelectorForSinglePixel();
4214 }
4215
4226 {
4227 mImpl->setNearestRounding(value);
4228 }
4229
4236 {
4237 return mImpl->getNearestRounding();
4238 }
4239
4257 void setCubicCoeff(float A) noexcept
4258 {
4259 mImpl->setCubicCoeff(A);
4260 }
4261
4267 float getCubicCoeff() const noexcept
4268 {
4269 return mImpl->getCubicCoeff();
4270 }
4271
4280 void setExcludeOutside(bool excludeFlag) noexcept
4281 {
4282 mImpl->setExcludeOutside(excludeFlag);
4283 }
4284
4290 bool getExcludeOutside() const noexcept
4291 {
4292 return mImpl->getExcludeOutside();
4293 }
4294
4295protected:
4296 virtual ~IResizeLayer() noexcept = default;
4297 apiv::VResizeLayer* mImpl;
4298};
4299
4303enum class LoopOutput : int32_t
4304{
4306 kLAST_VALUE = 0,
4307
4309 kCONCATENATE = 1,
4310
4312 kREVERSE = 2
4313};
4314
4320template <>
4321constexpr inline int32_t EnumMax<LoopOutput>() noexcept
4322{
4323 return 3;
4324}
4325
4329enum class TripLimit : int32_t
4330{
4331
4332 kCOUNT = 0,
4333 kWHILE = 1
4334};
4335
4341template <>
4342constexpr inline int32_t EnumMax<TripLimit>() noexcept
4343{
4344 return 2;
4345}
4346
4347class ILoop;
4348
4363{
4364public:
4368 ILoop* getLoop() const noexcept
4369 {
4370 return mBoundary->getLoop();
4371 }
4372
4373protected:
4374 virtual ~ILoopBoundaryLayer() noexcept = default;
4375 apiv::VLoopBoundaryLayer* mBoundary;
4376};
4377
4386{
4387public:
4392 {
4393 return mBoundary->getConditional();
4394 }
4395
4396protected:
4397 virtual ~IIfConditionalBoundaryLayer() noexcept = default;
4398 apiv::VConditionalBoundaryLayer* mBoundary;
4399};
4400
4407{
4408public:
4409protected:
4410 virtual ~IConditionLayer() noexcept = default;
4411 apiv::VConditionLayer* mImpl;
4412};
4413
4422{
4423public:
4424protected:
4425 virtual ~IIfConditionalOutputLayer() noexcept = default;
4426 apiv::VConditionalOutputLayer* mImpl;
4427};
4428
4435{
4436public:
4437protected:
4438 virtual ~IIfConditionalInputLayer() noexcept = default;
4439 apiv::VConditionalInputLayer* mImpl;
4440};
4441
4466{
4467public:
4478 {
4479 return mImpl->setCondition(condition);
4480 }
4481
4493 IIfConditionalOutputLayer* addOutput(ITensor& trueSubgraphOutput, ITensor& falseSubgraphOutput) noexcept
4494 {
4495 return mImpl->addOutput(trueSubgraphOutput, falseSubgraphOutput);
4496 }
4497
4506 {
4507 return mImpl->addInput(input);
4508 }
4509
4520 void setName(char const* name) noexcept
4521 {
4522 mImpl->setName(name);
4523 }
4524
4530 char const* getName() const noexcept
4531 {
4532 return mImpl->getName();
4533 }
4534
4535protected:
4536 virtual ~IIfConditional() noexcept = default;
4537 apiv::VIfConditional* mImpl;
4538};
4539
4548{
4549public:
4555 //
4568 using ILayer::setInput;
4569
4570protected:
4571 virtual ~IRecurrenceLayer() noexcept = default;
4572 apiv::VRecurrenceLayer* mImpl;
4573};
4574
4595{
4596public:
4600 LoopOutput getLoopOutput() const noexcept
4601 {
4602 return mImpl->getLoopOutput();
4603 }
4604
4617 void setAxis(int32_t axis) noexcept
4618 {
4619 mImpl->setAxis(axis);
4620 }
4621
4625 int32_t getAxis() const noexcept
4626 {
4627 return mImpl->getAxis();
4628 }
4629
4635 //
4650 using ILayer::setInput;
4651
4652protected:
4653 virtual ~ILoopOutputLayer() noexcept = default;
4654 apiv::VLoopOutputLayer* mImpl;
4655};
4656
4669{
4670public:
4674 TripLimit getTripLimit() const noexcept
4675 {
4676 return mImpl->getTripLimit();
4677 }
4678
4679protected:
4680 virtual ~ITripLimitLayer() noexcept = default;
4681 apiv::VTripLimitLayer* mImpl;
4682};
4683
4695{
4696public:
4700 void setAxis(int32_t axis) noexcept
4701 {
4702 mImpl->setAxis(axis);
4703 }
4704
4708 int32_t getAxis() const noexcept
4709 {
4710 return mImpl->getAxis();
4711 }
4712
4722 void setReverse(bool reverse) noexcept
4723 {
4724 mImpl->setReverse(reverse);
4725 }
4726
4732 bool getReverse() const noexcept
4733 {
4734 return mImpl->getReverse();
4735 }
4736
4737protected:
4738 virtual ~IIteratorLayer() noexcept = default;
4739 apiv::VIteratorLayer* mImpl;
4740};
4741
4751class ILoop : public INoCopy
4752{
4753public:
4760 IRecurrenceLayer* addRecurrence(ITensor& initialValue) noexcept
4761 {
4762 return mImpl->addRecurrence(initialValue);
4763 }
4764
4782 {
4783 return mImpl->addTripLimit(tensor, limit);
4784 }
4785
4794 IIteratorLayer* addIterator(ITensor& tensor, int32_t axis = 0, bool reverse = false) noexcept
4795 {
4796 return mImpl->addIterator(tensor, axis, reverse);
4797 }
4798
4807 ILoopOutputLayer* addLoopOutput(ITensor& tensor, LoopOutput outputKind, int32_t axis = 0) noexcept
4808 {
4809 return mImpl->addLoopOutput(tensor, outputKind, axis);
4810 }
4811
4822 void setName(char const* name) noexcept
4823 {
4824 mImpl->setName(name);
4825 }
4826
4832 char const* getName() const noexcept
4833 {
4834 return mImpl->getName();
4835 }
4836
4837protected:
4838 virtual ~ILoop() noexcept = default;
4839 apiv::VLoop* mImpl;
4840};
4841
4854class ISelectLayer : public ILayer
4855{
4856protected:
4857 virtual ~ISelectLayer() noexcept = default;
4858 apiv::VSelectLayer* mImpl;
4859};
4860
4877{
4878public:
4887 void setMessage(char const* message) noexcept
4888 {
4889 mImpl->setMessage(message);
4890 }
4891
4897 char const* getMessage() const noexcept
4898 {
4899 return mImpl->getMessage();
4900 }
4901
4902protected:
4903 virtual ~IAssertionLayer() noexcept = default;
4904
4905 apiv::VAssertionLayer* mImpl;
4906};
4907
4915enum class FillOperation : int32_t
4916{
4932 kLINSPACE = 0,
4933
4935 kRANDOM_UNIFORM = 1,
4936
4938 kRANDOM_NORMAL = 2
4939};
4940
4946template <>
4947constexpr inline int32_t EnumMax<FillOperation>() noexcept
4948{
4949 return 3;
4950}
4951
4987class IFillLayer : public ILayer
4988{
4989public:
4998 //
4999 void setDimensions(Dims const& dimensions) noexcept
5000 {
5001 mImpl->setDimensions(dimensions);
5002 }
5003
5014 Dims getDimensions() const noexcept
5015 {
5016 return mImpl->getDimensions();
5017 }
5018
5024 void setOperation(FillOperation op) noexcept
5025 {
5026 mImpl->setOperation(op);
5027 }
5028
5035 {
5036 return mImpl->getOperation();
5037 }
5038
5052 //
5053 void setAlpha(double alpha) noexcept
5054 {
5055 mImpl->setAlpha(alpha);
5056 }
5057
5068 double getAlpha() const noexcept
5069 {
5070 return mImpl->getAlpha();
5071 }
5072
5087 void setBeta(double beta) noexcept
5088 {
5089 mImpl->setBeta(beta);
5090 }
5091
5102 double getBeta() const noexcept
5103 {
5104 return mImpl->getBeta();
5105 }
5106
5147 using ILayer::setInput;
5148
5162 //
5163 void setAlphaInt64(int64_t alpha) noexcept
5164 {
5165 mImpl->setAlphaInt64(alpha);
5166 }
5167
5178 int64_t getAlphaInt64() const noexcept
5179 {
5180 return mImpl->getAlphaInt64();
5181 }
5182
5197 void setBetaInt64(int64_t beta) noexcept
5198 {
5199 mImpl->setBetaInt64(beta);
5200 }
5201
5212 int64_t getBetaInt64() const noexcept
5213 {
5214 return mImpl->getBetaInt64();
5215 }
5216
5220 bool isAlphaBetaInt64() const noexcept
5221 {
5222 return mImpl->isAlphaBetaInt64();
5223 }
5224
5237 void setToType(DataType toType) noexcept
5238 {
5239 mImpl->setToType(toType);
5240 }
5241
5249 DataType getToType() const noexcept
5250 {
5251 return mImpl->getToType();
5252 }
5253
5254protected:
5255 virtual ~IFillLayer() noexcept = default;
5256 apiv::VFillLayer* mImpl;
5257};
5258
5333{
5334public:
5343 int32_t getAxis() const noexcept
5344 {
5345 return mImpl->getAxis();
5346 }
5354 void setAxis(int32_t axis) noexcept
5355 {
5356 mImpl->setAxis(axis);
5357 }
5358
5370 void setToType(DataType toType) noexcept
5371 {
5372 mImpl->setToType(toType);
5373 }
5374
5382 DataType getToType() const noexcept
5383 {
5384 return mImpl->getToType();
5385 }
5386
5387protected:
5388 virtual ~IQuantizeLayer() noexcept = default;
5389 apiv::VQuantizeLayer* mImpl;
5390};
5391
5463{
5464public:
5473 int32_t getAxis() const noexcept
5474 {
5475 return mImpl->getAxis();
5476 }
5484 void setAxis(int32_t axis) noexcept
5485 {
5486 mImpl->setAxis(axis);
5487 }
5488
5500 void setToType(DataType toType) noexcept
5501 {
5502 mImpl->setToType(toType);
5503 }
5504
5512 DataType getToType() const noexcept
5513 {
5514 return mImpl->getToType();
5515 }
5516
5517protected:
5518 virtual ~IDequantizeLayer() noexcept = default;
5519 apiv::VDequantizeLayer* mImpl;
5520};
5521
5558class IEinsumLayer : public ILayer
5559{
5560public:
5570 bool setEquation(char const* equation) noexcept
5571 {
5572 return mImpl->setEquation(equation);
5573 }
5574
5580 char const* getEquation() const noexcept
5581 {
5582 return mImpl->getEquation();
5583 }
5584
5585protected:
5586 virtual ~IEinsumLayer() noexcept = default;
5587 apiv::VEinsumLayer* mImpl;
5588};
5589
5597enum class ScatterMode : int32_t
5598{
5599 kELEMENT = 0,
5600 kND = 1,
5601};
5602
5608template <>
5609constexpr inline int32_t EnumMax<ScatterMode>() noexcept
5610{
5611 return 2;
5612}
5613
5670class IScatterLayer : public ILayer
5671{
5672public:
5678 void setMode(ScatterMode mode) noexcept
5679 {
5680 mImpl->setMode(mode);
5681 }
5682
5688 ScatterMode getMode() const noexcept
5689 {
5690 return mImpl->getMode();
5691 }
5692
5698 void setAxis(int32_t axis) noexcept
5699 {
5700 mImpl->setAxis(axis);
5701 }
5702
5706 int32_t getAxis() const noexcept
5707 {
5708 return mImpl->getAxis();
5709 }
5710
5711protected:
5712 apiv::VScatterLayer* mImpl;
5713 virtual ~IScatterLayer() noexcept = default;
5714}; // class IScatterLayer
5715
5742class IOneHotLayer : public ILayer
5743{
5744public:
5750 void setAxis(int32_t axis) noexcept
5751 {
5752 mImpl->setAxis(axis);
5753 }
5754
5758 int32_t getAxis() const noexcept
5759 {
5760 return mImpl->getAxis();
5761 }
5762
5763protected:
5764 apiv::VOneHotLayer* mImpl;
5765};
5766
5779{
5780public:
5787 {
5788 mImpl->setInterpolationMode(mode);
5789 }
5790
5799 {
5800 return mImpl->getInterpolationMode();
5801 }
5802
5808 void setAlignCorners(bool alignCorners) noexcept
5809 {
5810 mImpl->setAlignCorners(alignCorners);
5811 }
5812
5820 bool getAlignCorners() const noexcept
5821 {
5822 return mImpl->getAlignCorners();
5823 }
5824
5832 bool setSampleMode(SampleMode mode) noexcept
5833 {
5834 return mImpl->setSampleMode(mode);
5835 }
5836
5844 SampleMode getSampleMode() const noexcept
5845 {
5846 return mImpl->getSampleMode();
5847 }
5848
5849protected:
5850 apiv::VGridSampleLayer* mImpl;
5851 virtual ~IGridSampleLayer() noexcept = default;
5852}; // class IGridSampleLayer
5853
5861enum class BoundingBoxFormat : int32_t
5862{
5864 kCORNER_PAIRS = 0,
5866 kCENTER_SIZES = 1
5867};
5868
5874template <>
5875constexpr inline int32_t EnumMax<BoundingBoxFormat>() noexcept
5876{
5877 return 2;
5878}
5879
5926class INMSLayer : public ILayer
5927{
5928public:
5939 {
5940 mImpl->setBoundingBoxFormat(fmt);
5941 }
5942
5951 {
5952 return mImpl->getBoundingBoxFormat();
5953 }
5954
5964 void setTopKBoxLimit(int32_t limit) noexcept
5965 {
5966 mImpl->setTopKBoxLimit(limit);
5967 }
5968
5974 int32_t getTopKBoxLimit() const noexcept
5975 {
5976 return mImpl->getTopKBoxLimit();
5977 }
5978
5997 using ILayer::setInput;
5998
5999protected:
6000 apiv::VNMSLayer* mImpl;
6001 virtual ~INMSLayer() noexcept = default;
6002}; // class INMSLayer
6003
6017{
6018public:
6027 void setBatchAxis(int32_t batchAxis) noexcept
6028 {
6029 mImpl->setBatchAxis(batchAxis);
6030 }
6031
6037 int32_t getBatchAxis() const noexcept
6038 {
6039 return mImpl->getBatchAxis();
6040 }
6041
6050 void setSequenceAxis(int32_t sequenceAxis) noexcept
6051 {
6052 mImpl->setSequenceAxis(sequenceAxis);
6053 }
6054
6060 int32_t getSequenceAxis() const noexcept
6061 {
6062 return mImpl->getSequenceAxis();
6063 }
6064
6065protected:
6066 apiv::VReverseSequenceLayer* mImpl;
6067 virtual ~IReverseSequenceLayer() noexcept = default;
6068}; // class IReverseSequenceLayer
6069
6089{
6090public:
6098 void setEpsilon(float eps) noexcept
6099 {
6100 return mImpl->setEpsilon(eps);
6101 }
6102
6108 float getEpsilon() const noexcept
6109 {
6110 return mImpl->getEpsilon();
6111 }
6112
6118 void setAxes(uint32_t axesMask) noexcept
6119 {
6120 return mImpl->setAxes(axesMask);
6121 }
6122
6128 uint32_t getAxes() const noexcept
6129 {
6130 return mImpl->getAxes();
6131 }
6132
6149 void setNbGroups(int64_t nbGroups) noexcept
6150 {
6151 return mImpl->setNbGroups(nbGroups);
6152 }
6153
6159 int64_t getNbGroups() const noexcept
6160 {
6161 return mImpl->getNbGroups();
6162 }
6163
6185 void setComputePrecision(DataType type) noexcept
6186 {
6187 return mImpl->setComputePrecision(type);
6188 }
6189
6196 {
6197 return mImpl->getComputePrecision();
6198 }
6199
6200protected:
6201 apiv::VNormalizationLayer* mImpl;
6202 virtual ~INormalizationLayer() noexcept = default;
6203};
6204
6223{
6224public:
6225 virtual ~INetworkDefinition() noexcept = default;
6226
6263 ITensor* addInput(char const* name, DataType type, Dims const& dimensions) noexcept
6264 {
6265 return mImpl->addInput(name, type, dimensions);
6266 }
6267
6277 void markOutput(ITensor& tensor) noexcept
6278 {
6279 mImpl->markOutput(tensor);
6280 }
6281
6295 bool markDebug(ITensor& tensor) noexcept
6296 {
6297 return mImpl->markDebug(tensor);
6298 }
6299
6311 bool unmarkDebug(ITensor& tensor) noexcept
6312 {
6313 return mImpl->unmarkDebug(tensor);
6314 }
6315
6321 bool isDebugTensor(nvinfer1::ITensor const& tensor) const noexcept
6322 {
6323 return mImpl->isDebugTensor(tensor);
6324 }
6325
6342 {
6343 return mImpl->addActivation(input, type);
6344 }
6345
6360 ILRNLayer* addLRN(ITensor& input, int64_t window, float alpha, float beta, float k) noexcept
6361 {
6362 return mImpl->addLRN(input, window, alpha, beta, k);
6363 }
6364
6386 IScaleLayer* addScale(ITensor& input, ScaleMode mode, Weights shift, Weights scale, Weights power) noexcept
6387 {
6388 return mImpl->addScale(input, mode, shift, scale, power);
6389 }
6390
6400 {
6401 return mImpl->addSoftMax(input);
6402 }
6403
6416 IConcatenationLayer* addConcatenation(ITensor* const* inputs, int32_t nbInputs) noexcept
6417 {
6418 return mImpl->addConcatenation(inputs, nbInputs);
6419 }
6420
6444 {
6445 return mImpl->addElementWise(input1, input2, op);
6446 }
6447
6465 IUnaryLayer* addUnary(ITensor& input, UnaryOperation operation) noexcept
6466 {
6467 return mImpl->addUnary(input, operation);
6468 }
6469
6480 {
6481 return mImpl->addShuffle(input);
6482 }
6483
6496 IOneHotLayer* addOneHot(ITensor& indices, ITensor& values, ITensor& depth, int32_t axis) noexcept
6497 {
6498 return mImpl->addOneHot(indices, values, depth, axis);
6499 }
6500
6508 int32_t getNbLayers() const noexcept
6509 {
6510 return mImpl->getNbLayers();
6511 }
6512
6522 ILayer* getLayer(int32_t index) const noexcept
6523 {
6524 return mImpl->getLayer(index);
6525 }
6526
6534 int32_t getNbInputs() const noexcept
6535 {
6536 return mImpl->getNbInputs();
6537 }
6538
6550 ITensor* getInput(int32_t index) const noexcept
6551 {
6552 return mImpl->getInput(index);
6553 }
6554
6564 int32_t getNbOutputs() const noexcept
6565 {
6566 return mImpl->getNbOutputs();
6567 }
6568
6580 ITensor* getOutput(int32_t index) const noexcept
6581 {
6582 return mImpl->getOutput(index);
6583 }
6584
6607 ITensor& input, ReduceOperation operation, uint32_t reduceAxes, bool keepDimensions) noexcept
6608 {
6609 return mImpl->addReduce(input, operation, reduceAxes, keepDimensions);
6610 }
6611
6639 ITopKLayer* addTopK(ITensor& input, TopKOperation op, int32_t k, uint32_t reduceAxes) noexcept
6640 {
6641 return mImpl->addTopK(input, op, k, reduceAxes);
6642 }
6643
6655 IGatherLayer* addGather(ITensor& data, ITensor& indices, int32_t axis) noexcept
6656 {
6657 return mImpl->addGather(data, indices, axis);
6658 }
6659
6671 IGatherLayer* addGatherV2(ITensor& data, ITensor& indices, GatherMode mode) noexcept
6672 {
6673 return mImpl->addGatherV2(data, indices, mode);
6674 }
6675
6691 {
6692 return mImpl->addRaggedSoftMax(input, bounds);
6693 }
6694
6712 ITensor& input0, MatrixOperation op0, ITensor& input1, MatrixOperation op1) noexcept
6713 {
6714 return mImpl->addMatrixMultiply(input0, op0, input1, op1);
6715 }
6716
6727 {
6728 return mImpl->addNonZero(input);
6729 }
6730
6750 IConstantLayer* addConstant(Dims const& dimensions, Weights weights) noexcept
6751 {
6752 return mImpl->addConstant(dimensions, weights);
6753 }
6754
6765 {
6766 return mImpl->addIdentity(input);
6767 }
6768
6779 ICastLayer* addCast(ITensor& input, DataType toType) noexcept
6780 {
6781 return mImpl->addCast(input, toType);
6782 }
6783
6794 void removeTensor(ITensor& tensor) noexcept
6795 {
6796 mImpl->removeTensor(tensor);
6797 }
6798
6806 void unmarkOutput(ITensor& tensor) noexcept
6807 {
6808 mImpl->unmarkOutput(tensor);
6809 }
6810
6825 IPluginV2Layer* addPluginV2(ITensor* const* inputs, int32_t nbInputs, IPluginV2& plugin) noexcept
6826 {
6827 return mImpl->addPluginV2(inputs, nbInputs, plugin);
6828 }
6829
6843 IPluginV3Layer* addPluginV3(ITensor* const* inputs, int32_t nbInputs, ITensor* const* shapeInputs,
6844 int32_t nbShapeInputs, IPluginV3& plugin) noexcept
6845 {
6846 return mImpl->addPluginV3(inputs, nbInputs, shapeInputs, nbShapeInputs, plugin);
6847 }
6848
6863 ISliceLayer* addSlice(ITensor& input, Dims const& start, Dims const& size, Dims const& stride) noexcept
6864 {
6865 return mImpl->addSlice(input, start, size, stride);
6866 }
6867
6887 void setName(char const* name) noexcept
6888 {
6889 mImpl->setName(name);
6890 }
6891
6901 char const* getName() const noexcept
6902 {
6903 return mImpl->getName();
6904 }
6905
6917 IShapeLayer* addShape(ITensor& input) noexcept
6918 {
6919 return mImpl->addShape(input);
6920 }
6921
6932 {
6933 return mImpl->hasImplicitBatchDimension();
6934 }
6935
6942 {
6943 return mImpl->getFlags();
6944 }
6945
6953 bool getFlag(NetworkDefinitionCreationFlag networkDefinitionCreationFlag) const noexcept
6954 {
6955 return mImpl->getFlag(networkDefinitionCreationFlag);
6956 }
6957
6970 bool markOutputForShapes(ITensor& tensor) noexcept
6971 {
6972 return mImpl->markOutputForShapes(tensor);
6973 }
6974
6982 bool unmarkOutputForShapes(ITensor& tensor) noexcept
6983 {
6984 return mImpl->unmarkOutputForShapes(tensor);
6985 }
6986
7001 {
7002 return mImpl->addParametricReLU(input, slope);
7003 }
7004
7023 ITensor& input, int64_t nbOutputMaps, Dims const& kernelSize, Weights kernelWeights, Weights biasWeights) noexcept
7024 {
7025 return mImpl->addConvolutionNd(input, nbOutputMaps, kernelSize, kernelWeights, biasWeights);
7026 }
7027
7042 IPoolingLayer* addPoolingNd(ITensor& input, PoolingType type, Dims const& windowSize) noexcept
7043 {
7044 return mImpl->addPoolingNd(input, type, windowSize);
7045 }
7046
7061 //
7065 ITensor& input, int64_t nbOutputMaps, Dims kernelSize, Weights kernelWeights, Weights biasWeights) noexcept
7066 {
7067 return mImpl->addDeconvolutionNd(input, nbOutputMaps, kernelSize, kernelWeights, biasWeights);
7068 }
7069
7102 ITensor& input, ScaleMode mode, Weights shift, Weights scale, Weights power, int32_t channelAxis) noexcept
7103 {
7104 return mImpl->addScaleNd(input, mode, shift, scale, power, channelAxis);
7105 }
7106
7119 {
7120 return mImpl->addResize(input);
7121 }
7122
7132 ILoop* addLoop() noexcept
7133 {
7134 return mImpl->addLoop();
7135 }
7136
7148 {
7149 return mImpl->addIfConditional();
7150 }
7151
7186 ISelectLayer* addSelect(ITensor& condition, ITensor& thenInput, ITensor& elseInput) noexcept
7187 {
7188 return mImpl->addSelect(condition, thenInput, elseInput);
7189 }
7190
7203 IAssertionLayer* addAssertion(ITensor& condition, char const* message) noexcept
7204 {
7205 return mImpl->addAssertion(condition, message);
7206 }
7207
7228 TRT_DEPRECATED IFillLayer* addFill(Dims const& dimensions, FillOperation op) noexcept
7229 {
7230 return mImpl->addFill(dimensions, op);
7231 }
7232
7254 IFillLayer* addFill(Dims const& dimensions, FillOperation op, DataType outputType) noexcept
7255 {
7256 return mImpl->addFillV2(dimensions, op, outputType);
7257 }
7258
7270 IPaddingLayer* addPaddingNd(ITensor& input, Dims const& prePadding, Dims const& postPadding) noexcept
7271 {
7272 return mImpl->addPaddingNd(input, prePadding, postPadding);
7273 }
7274
7294 bool setWeightsName(Weights weights, char const* name) noexcept
7295 {
7296 return mImpl->setWeightsName(weights, name);
7297 }
7298
7310 //
7313 void setErrorRecorder(IErrorRecorder* recorder) noexcept
7314 {
7315 mImpl->setErrorRecorder(recorder);
7316 }
7317
7329 {
7330 return mImpl->getErrorRecorder();
7331 }
7332
7350 {
7351 return mImpl->addDequantize(input, scale);
7352 }
7353
7370 IDequantizeLayer* addDequantize(ITensor& input, ITensor& scale, DataType outputType) noexcept
7371 {
7372 return mImpl->addDequantizeV2(input, scale, outputType);
7373 }
7374
7390 IScatterLayer* addScatter(ITensor& data, ITensor& indices, ITensor& updates, ScatterMode mode) noexcept
7391 {
7392 return mImpl->addScatter(data, indices, updates, mode);
7393 }
7394
7412 {
7413 return mImpl->addQuantize(input, scale);
7414 }
7415
7432 IQuantizeLayer* addQuantize(ITensor& input, ITensor& scale, DataType outputType) noexcept
7433 {
7434 return mImpl->addQuantizeV2(input, scale, outputType);
7435 }
7436
7447 IEinsumLayer* addEinsum(ITensor* const* inputs, int32_t nbInputs, char const* equation) noexcept
7448 {
7449 return mImpl->addEinsum(inputs, nbInputs, equation);
7450 }
7451
7466 {
7467 return mImpl->addGridSample(input, grid);
7468 }
7469
7483 INMSLayer* addNMS(ITensor& boxes, ITensor& scores, ITensor& maxOutputBoxesPerClass) noexcept
7484 {
7485 return mImpl->addNMS(boxes, scores, maxOutputBoxesPerClass);
7486 }
7487
7501 {
7502 return mImpl->addReverseSequence(input, sequenceLens);
7503 }
7504
7526 INormalizationLayer* addNormalization(ITensor& input, ITensor& scale, ITensor& bias, uint32_t axesMask) noexcept
7527 {
7528 return mImpl->addNormalization(input, scale, bias, axesMask);
7529 }
7530
7537 virtual IBuilder& getBuilder() const noexcept
7538 {
7539 return mImpl->getBuilder();
7540 }
7541
7542protected:
7543 apiv::VNetworkDefinition* mImpl;
7544};
7545
7553enum class CalibrationAlgoType : int32_t
7554{
7559};
7560
7566template <>
7567constexpr inline int32_t EnumMax<CalibrationAlgoType>() noexcept
7568{
7569 return 4;
7570}
7571
7586{
7587public:
7595 TRT_DEPRECATED virtual int32_t getBatchSize() const noexcept = 0;
7596
7611 virtual bool getBatch(void* bindings[], char const* names[], int32_t nbBindings) noexcept = 0;
7612
7627 virtual void const* readCalibrationCache(std::size_t& length) noexcept = 0;
7628
7637 virtual void writeCalibrationCache(void const* ptr, std::size_t length) noexcept = 0;
7638
7644 virtual CalibrationAlgoType getAlgorithm() noexcept = 0;
7645
7646 ~IInt8Calibrator() noexcept override = default;
7647};
7648
7649namespace v_1_0
7650{
7652{
7653public:
7657 InterfaceInfo getInterfaceInfo() const noexcept override
7658 {
7659 return InterfaceInfo{"IInt8EntropyCalibrator", 1, 0};
7660 }
7661
7666 {
7668 }
7669
7670 ~IInt8EntropyCalibrator() noexcept override = default;
7671};
7672} // namespace v_1_0
7673
7688
7689namespace v_1_0
7690{
7692{
7693public:
7697 InterfaceInfo getInterfaceInfo() const noexcept override
7698 {
7699 return InterfaceInfo{"IInt8EntropyCalibrator2", 1, 0};
7700 }
7701
7706 {
7708 }
7709
7710 ~IInt8EntropyCalibrator2() noexcept override = default;
7711};
7712} // namespace v_1_0
7713
7728
7729namespace v_1_0
7730{
7732{
7733public:
7737 InterfaceInfo getInterfaceInfo() const noexcept override
7738 {
7739 return InterfaceInfo{"IInt8MinMaxCalibrator", 1, 0};
7740 }
7741
7746 {
7748 }
7749
7750 ~IInt8MinMaxCalibrator() noexcept override = default;
7751};
7752} // namespace v_1_0
7753
7767
7768namespace v_1_0
7769{
7771{
7772public:
7776 InterfaceInfo getInterfaceInfo() const noexcept override
7777 {
7778 return InterfaceInfo{"IInt8Calibrator", 1, 0};
7779 }
7780
7785 {
7787 }
7788
7795 virtual double getQuantile() const noexcept = 0;
7796
7803 virtual double getRegressionCutoff() const noexcept = 0;
7804
7817 virtual void const* readHistogramCache(std::size_t& length) noexcept = 0;
7818
7827 virtual void writeHistogramCache(void const* ptr, std::size_t length) noexcept = 0;
7828
7829 ~IInt8LegacyCalibrator() noexcept override = default;
7830};
7831} // namespace v_1_0
7832
7847
7859{
7860public:
7866 DataType getDataType() const noexcept
7867 {
7868 return mImpl->getDataType();
7869 }
7870
7877 Dims getStrides() const noexcept
7878 {
7879 return mImpl->getStrides();
7880 }
7881
7887 int64_t getVectorizedDim() const noexcept
7888 {
7889 return mImpl->getVectorizedDim();
7890 }
7891
7898 int64_t getComponentsPerElement() const noexcept
7899 {
7900 return mImpl->getComponentsPerElement();
7901 }
7902
7903protected:
7904 virtual ~IAlgorithmIOInfo() noexcept = default;
7905 apiv::VAlgorithmIOInfo* mImpl;
7906};
7907
7920{
7921public:
7925 int64_t getImplementation() const noexcept
7926 {
7927 return mImpl->getImplementation();
7928 }
7929
7933 int64_t getTactic() const noexcept
7934 {
7935 return mImpl->getTactic();
7936 }
7937
7938protected:
7939 virtual ~IAlgorithmVariant() noexcept = default;
7940 apiv::VAlgorithmVariant* mImpl;
7941};
7942
7952{
7953public:
7959 char const* getName() const noexcept
7960 {
7961 return mImpl->getName();
7962 }
7963
7971 Dims getDimensions(int32_t index, OptProfileSelector select) const noexcept
7972 {
7973 return mImpl->getDimensions(index, select);
7974 }
7975
7979 int32_t getNbInputs() const noexcept
7980 {
7981 return mImpl->getNbInputs();
7982 }
7983
7987 int32_t getNbOutputs() const noexcept
7988 {
7989 return mImpl->getNbOutputs();
7990 }
7991
7992protected:
7993 virtual ~IAlgorithmContext() noexcept = default;
7994 apiv::VAlgorithmContext* mImpl;
7995};
7996
8008class IAlgorithm : public INoCopy
8009{
8010public:
8015 {
8016 return mImpl->getAlgorithmVariant();
8017 }
8018
8022 float getTimingMSec() const noexcept
8023 {
8024 return mImpl->getTimingMSec();
8025 }
8026
8030 std::size_t getWorkspaceSize() const noexcept
8031 {
8032 return mImpl->getWorkspaceSize();
8033 }
8034
8044 IAlgorithmIOInfo const* getAlgorithmIOInfoByIndex(int32_t index) const noexcept
8045 {
8046 return mImpl->getAlgorithmIOInfoByIndex(index);
8047 }
8048
8049protected:
8050 virtual ~IAlgorithm() noexcept = default;
8051 apiv::VAlgorithm* mImpl;
8052}; // IAlgorithm
8053
8054namespace v_1_0
8055{
8057{
8058public:
8062 InterfaceInfo getInterfaceInfo() const noexcept override
8063 {
8064 return InterfaceInfo{"IAlgorithmSelector", 1, 0};
8065 }
8080 virtual int32_t selectAlgorithms(IAlgorithmContext const& context, IAlgorithm const* const* choices,
8081 int32_t nbChoices, int32_t* selection) noexcept = 0;
8082
8093 virtual void reportAlgorithms(IAlgorithmContext const* const* algoContexts, IAlgorithm const* const* algoChoices,
8094 int32_t nbAlgorithms) noexcept = 0;
8095
8096 virtual ~IAlgorithmSelector() noexcept = default;
8097};
8098} // namespace v_1_0
8099
8111
8118using QuantizationFlags = uint32_t;
8119
8129enum class QuantizationFlag : int32_t
8130{
8135};
8136
8142template <>
8143constexpr inline int32_t EnumMax<QuantizationFlag>() noexcept
8144{
8145 return 1;
8146}
8147
8154using BuilderFlags = uint32_t;
8155
8163enum class BuilderFlag : int32_t
8164{
8166 kFP16 = 0,
8167
8169 kINT8 = 1,
8170
8172 kDEBUG = 2,
8173
8175 kGPU_FALLBACK = 3,
8176
8178 kREFIT = 4,
8179
8182
8186 kTF32 = 6,
8187
8189 kSPARSE_WEIGHTS = 7,
8190
8197 kSAFETY_SCOPE = 8,
8198
8201
8205
8209 kDIRECT_IO = 11,
8210
8213
8220
8226
8232 kFP8 = 15,
8233
8237
8240 kBF16 = 17,
8241
8247
8249 kSTRIP_PLAN = 19,
8250
8253
8260 kREFIT_IDENTICAL = 20,
8261
8287 kWEIGHT_STREAMING = 21,
8288
8290 kINT4 = 22,
8291
8292};
8293
8299template <>
8300constexpr inline int32_t EnumMax<BuilderFlag>() noexcept
8301{
8302 return 23;
8303}
8304
8318class ITimingCache : public INoCopy
8319{
8320public:
8321 virtual ~ITimingCache() noexcept = default;
8322
8332 nvinfer1::IHostMemory* serialize() const noexcept
8333 {
8334 return mImpl->serialize();
8335 }
8336
8356 bool combine(ITimingCache const& inputCache, bool ignoreMismatch) noexcept
8357 {
8358 return mImpl->combine(inputCache, ignoreMismatch);
8359 }
8360
8366 bool reset() noexcept
8367 {
8368 return mImpl->reset();
8369 }
8370
8371protected:
8372 apiv::VTimingCache* mImpl;
8373};
8374
8382enum class MemoryPoolType : int32_t
8383{
8390 kWORKSPACE = 0,
8391
8399
8405 kDLA_LOCAL_DRAM = 2,
8406
8412 kDLA_GLOBAL_DRAM = 3,
8413
8421 kTACTIC_DRAM = 4,
8422
8436};
8437
8443template <>
8444constexpr inline int32_t EnumMax<MemoryPoolType>() noexcept
8445{
8446 return 6;
8447}
8448
8457enum class PreviewFeature : int32_t
8458{
8465};
8466namespace impl
8467{
8473template <>
8475{
8476 static constexpr int32_t kVALUE = 1;
8477};
8478} // namespace impl
8479
8490enum class HardwareCompatibilityLevel : int32_t
8491{
8494 kNONE = 0,
8495
8505 kAMPERE_PLUS = 1,
8506};
8507
8508namespace impl
8509{
8515template <>
8517{
8518 static constexpr int32_t kVALUE = 2;
8519};
8520} // namespace impl
8521
8522namespace v_1_0
8523{
8525{
8526public:
8527 IProgressMonitor() = default;
8528 virtual ~IProgressMonitor() noexcept = default;
8529
8533 InterfaceInfo getInterfaceInfo() const noexcept override
8534 {
8535 return InterfaceInfo{"IProgressMonitor", 1, 0};
8536 }
8537
8557 virtual void phaseStart(char const* phaseName, char const* parentPhase, int32_t nbSteps) noexcept = 0;
8558
8571 virtual bool stepComplete(char const* phaseName, int32_t step) noexcept = 0;
8572
8584 virtual void phaseFinish(char const* phaseName) noexcept = 0;
8585
8586}; // class IProgressMonitor
8587} // namespace v_1_0
8588
8609
8618{
8619public:
8620 virtual ~IBuilderConfig() noexcept = default;
8621
8630 virtual void setAvgTimingIterations(int32_t avgTiming) noexcept
8631 {
8632 mImpl->setAvgTimingIterations(avgTiming);
8633 }
8634
8642 int32_t getAvgTimingIterations() const noexcept
8643 {
8644 return mImpl->getAvgTimingIterations();
8645 }
8646
8655 void setEngineCapability(EngineCapability capability) noexcept
8656 {
8657 mImpl->setEngineCapability(capability);
8658 }
8659
8668 {
8669 return mImpl->getEngineCapability();
8670 }
8671
8680 {
8681 mImpl->setInt8Calibrator(calibrator);
8682 }
8683
8690 {
8691 return mImpl->getInt8Calibrator();
8692 }
8693
8706 void setFlags(BuilderFlags builderFlags) noexcept
8707 {
8708 mImpl->setFlags(builderFlags);
8709 }
8710
8718 BuilderFlags getFlags() const noexcept
8719 {
8720 return mImpl->getFlags();
8721 }
8722
8730 void clearFlag(BuilderFlag builderFlag) noexcept
8731 {
8732 mImpl->clearFlag(builderFlag);
8733 }
8734
8742 void setFlag(BuilderFlag builderFlag) noexcept
8743 {
8744 mImpl->setFlag(builderFlag);
8745 }
8746
8754 bool getFlag(BuilderFlag builderFlag) const noexcept
8755 {
8756 return mImpl->getFlag(builderFlag);
8757 }
8758
8771 void setDeviceType(ILayer const* layer, DeviceType deviceType) noexcept
8772 {
8773 mImpl->setDeviceType(layer, deviceType);
8774 }
8775
8781 DeviceType getDeviceType(ILayer const* layer) const noexcept
8782 {
8783 return mImpl->getDeviceType(layer);
8784 }
8785
8793 bool isDeviceTypeSet(ILayer const* layer) const noexcept
8794 {
8795 return mImpl->isDeviceTypeSet(layer);
8796 }
8797
8803 void resetDeviceType(ILayer const* layer) noexcept
8804 {
8805 mImpl->resetDeviceType(layer);
8806 }
8807
8813 bool canRunOnDLA(ILayer const* layer) const noexcept
8814 {
8815 return mImpl->canRunOnDLA(layer);
8816 }
8817
8829 void setDLACore(int32_t dlaCore) noexcept
8830 {
8831 mImpl->setDLACore(dlaCore);
8832 }
8833
8839 int32_t getDLACore() const noexcept
8840 {
8841 return mImpl->getDLACore();
8842 }
8843
8850 void setDefaultDeviceType(DeviceType deviceType) noexcept
8851 {
8852 mImpl->setDefaultDeviceType(deviceType);
8853 }
8854
8861 {
8862 return mImpl->getDefaultDeviceType();
8863 }
8864
8870 void reset() noexcept
8871 {
8872 mImpl->reset();
8873 }
8874
8882 void setProfileStream(const cudaStream_t stream) noexcept
8883 {
8884 return mImpl->setProfileStream(stream);
8885 }
8886
8894 cudaStream_t getProfileStream() const noexcept
8895 {
8896 return mImpl->getProfileStream();
8897 }
8898
8911 int32_t addOptimizationProfile(IOptimizationProfile const* profile) noexcept
8912 {
8913 return mImpl->addOptimizationProfile(profile);
8914 }
8915
8924 int32_t getNbOptimizationProfiles() const noexcept
8925 {
8926 return mImpl->getNbOptimizationProfiles();
8927 }
8928
8937 {
8938 mImpl->setProfilingVerbosity(verbosity);
8939 }
8940
8950 {
8951 return mImpl->getProfilingVerbosity();
8952 }
8953
8959 {
8960 mImpl->setAlgorithmSelector(selector);
8961 }
8962
8967 {
8968 return mImpl->getAlgorithmSelector();
8969 }
8970
8985 {
8986 return mImpl->setCalibrationProfile(profile);
8987 }
8988
8997 {
8998 return mImpl->getCalibrationProfile();
8999 }
9000
9014 {
9015 mImpl->setQuantizationFlags(flags);
9016 }
9017
9026 {
9027 return mImpl->getQuantizationFlags();
9028 }
9029
9038 {
9039 mImpl->clearQuantizationFlag(flag);
9040 }
9041
9050 {
9051 mImpl->setQuantizationFlag(flag);
9052 }
9053
9061 bool getQuantizationFlag(QuantizationFlag flag) const noexcept
9062 {
9063 return mImpl->getQuantizationFlag(flag);
9064 }
9065
9083 bool setTacticSources(TacticSources tacticSources) noexcept
9084 {
9085 return mImpl->setTacticSources(tacticSources);
9086 }
9087
9099 {
9100 return mImpl->getTacticSources();
9101 }
9102
9117 nvinfer1::ITimingCache* createTimingCache(void const* blob, std::size_t size) const noexcept
9118 {
9119 return mImpl->createTimingCache(blob, size);
9120 }
9121
9140 bool setTimingCache(ITimingCache const& cache, bool ignoreMismatch) noexcept
9141 {
9142 return mImpl->setTimingCache(cache, ignoreMismatch);
9143 }
9144
9151 {
9152 return mImpl->getTimingCache();
9153 }
9154
9182 void setMemoryPoolLimit(MemoryPoolType pool, std::size_t poolSize) noexcept
9183 {
9184 mImpl->setMemoryPoolLimit(pool, poolSize);
9185 }
9186
9201 std::size_t getMemoryPoolLimit(MemoryPoolType pool) const noexcept
9202 {
9203 return mImpl->getMemoryPoolLimit(pool);
9204 }
9205
9219 void setPreviewFeature(PreviewFeature feature, bool enable) noexcept
9220 {
9221 mImpl->setPreviewFeature(feature, enable);
9222 }
9223
9233 bool getPreviewFeature(PreviewFeature feature) const noexcept
9234 {
9235 return mImpl->getPreviewFeature(feature);
9236 }
9237
9266 void setBuilderOptimizationLevel(int32_t level) noexcept
9267 {
9268 mImpl->setBuilderOptimizationLevel(level);
9269 }
9270
9279 {
9280 return mImpl->getBuilderOptimizationLevel();
9281 }
9282
9295 void setHardwareCompatibilityLevel(HardwareCompatibilityLevel hardwareCompatibilityLevel) noexcept
9296 {
9297 mImpl->setHardwareCompatibilityLevel(hardwareCompatibilityLevel);
9298 }
9299
9309 {
9310 return mImpl->getHardwareCompatibilityLevel();
9311 }
9312
9321 void setPluginsToSerialize(char const* const* paths, int32_t nbPaths) noexcept
9322 {
9323 mImpl->setPluginsToSerialize(paths, nbPaths);
9324 }
9325
9334 char const* getPluginToSerialize(int32_t index) const noexcept
9335 {
9336 return mImpl->getPluginToSerialize(index);
9337 }
9338
9344 int32_t getNbPluginsToSerialize() const noexcept
9345 {
9346 return mImpl->getNbPluginsToSerialize();
9347 }
9348
9373 void setMaxAuxStreams(int32_t nbStreams) noexcept
9374 {
9375 mImpl->setMaxAuxStreams(nbStreams);
9376 }
9377
9383 int32_t getMaxAuxStreams() const noexcept
9384 {
9385 return mImpl->getMaxAuxStreams();
9386 }
9387
9399 void setProgressMonitor(IProgressMonitor* monitor) noexcept
9400 {
9401 return mImpl->setProgressMonitor(monitor);
9402 }
9403
9410 {
9411 return mImpl->getProgressMonitor();
9412 }
9413
9414protected:
9415 apiv::VBuilderConfig* mImpl;
9416};
9417
9426
9436{
9441
9446 kSTRONGLY_TYPED = 1,
9447};
9448
9454template <>
9455constexpr inline int32_t EnumMax<NetworkDefinitionCreationFlag>() noexcept
9456{
9457 return 2;
9458}
9459
9467class IBuilder : public INoCopy
9468{
9469public:
9470 virtual ~IBuilder() noexcept = default;
9471
9475 bool platformHasFastFp16() const noexcept
9476 {
9477 return mImpl->platformHasFastFp16();
9478 }
9479
9483 bool platformHasFastInt8() const noexcept
9484 {
9485 return mImpl->platformHasFastInt8();
9486 }
9487
9495 int32_t getMaxDLABatchSize() const noexcept
9496 {
9497 return mImpl->getMaxDLABatchSize();
9498 }
9499
9503 int32_t getNbDLACores() const noexcept
9504 {
9505 return mImpl->getNbDLACores();
9506 }
9507
9520 void setGpuAllocator(IGpuAllocator* allocator) noexcept
9521 {
9522 mImpl->setGpuAllocator(allocator);
9523 }
9524
9531 {
9532 return mImpl->createBuilderConfig();
9533 }
9534
9553 {
9554 return mImpl->createNetworkV2(flags);
9555 }
9556
9568 {
9569 return mImpl->createOptimizationProfile();
9570 }
9571
9586 void setErrorRecorder(IErrorRecorder* recorder) noexcept
9587 {
9588 mImpl->setErrorRecorder(recorder);
9589 }
9590
9602 {
9603 return mImpl->getErrorRecorder();
9604 }
9605
9609 void reset() noexcept
9610 {
9611 mImpl->reset();
9612 }
9613
9617 bool platformHasTf32() const noexcept
9618 {
9619 return mImpl->platformHasTf32();
9620 }
9621
9637 {
9638 return mImpl->buildSerializedNetwork(network, config);
9639 }
9640
9658 bool isNetworkSupported(INetworkDefinition const& network, IBuilderConfig const& config) const noexcept
9659 {
9660 return mImpl->isNetworkSupported(network, config);
9661 }
9662
9668 ILogger* getLogger() const noexcept
9669 {
9670 return mImpl->getLogger();
9671 }
9672
9684 bool setMaxThreads(int32_t maxThreads) noexcept
9685 {
9686 return mImpl->setMaxThreads(maxThreads);
9687 }
9688
9698 int32_t getMaxThreads() const noexcept
9699 {
9700 return mImpl->getMaxThreads();
9701 }
9702
9709 {
9710 return mImpl->getPluginRegistry();
9711 }
9712
9713protected:
9714 apiv::VBuilder* mImpl;
9715};
9716
9717} // namespace nvinfer1
9718
9723extern "C" TENSORRTAPI void* createInferBuilder_INTERNAL(void* logger, int32_t version) noexcept;
9724
9725namespace nvinfer1
9726{
9727namespace
9728{
9729
9737inline IBuilder* createInferBuilder(ILogger& logger) noexcept
9738{
9739 return static_cast<IBuilder*>(createInferBuilder_INTERNAL(&logger, NV_TENSORRT_VERSION));
9740}
9741
9742} // namespace
9743
9757 nvinfer1::EngineCapability capability) noexcept;
9758
9759namespace safe
9760{
9762class IPluginRegistry;
9763} // namespace safe
9764
9773 nvinfer1::EngineCapability capability) noexcept;
9774
9775} // namespace nvinfer1
9776
9777#endif // NV_INFER_H
#define TENSORRTAPI
Definition: NvInferRuntimeBase.h:59
#define NV_TENSORRT_VERSION
Definition: NvInferRuntimeBase.h:93
#define TRT_DEPRECATED
Definition: NvInferRuntimeBase.h:45
#define TRT_DEPRECATED_ENUM
Definition: NvInferRuntimeBase.h:46
Definition: NvInferRuntimeBase.h:202
static constexpr int32_t MAX_DIMS
The maximum rank (number of dimensions) supported for a tensor.
Definition: NvInferRuntimeBase.h:205
An Activation layer in a network definition.
Definition: NvInfer.h:1342
void setBeta(float beta) noexcept
Set the beta parameter (must be finite).
Definition: NvInfer.h:1390
void setActivationType(ActivationType type) noexcept
Set the type of activation to be performed.
Definition: NvInfer.h:1351
ActivationType getActivationType() const noexcept
Get the type of activation to be performed.
Definition: NvInfer.h:1361
float getAlpha() const noexcept
Get the alpha parameter.
Definition: NvInfer.h:1399
virtual ~IActivationLayer() noexcept=default
float getBeta() const noexcept
Get the beta parameter.
Definition: NvInfer.h:1408
void setAlpha(float alpha) noexcept
Set the alpha parameter (must be finite).
Definition: NvInfer.h:1376
Describes the context and requirements, that could be fulfilled by one or more instances of IAlgorith...
Definition: NvInfer.h:7952
int32_t getNbOutputs() const noexcept
Return number of outputs of the algorithm.
Definition: NvInfer.h:7987
int32_t getNbInputs() const noexcept
Return number of inputs of the algorithm.
Definition: NvInfer.h:7979
char const * getName() const noexcept
Return name of the algorithm node.
Definition: NvInfer.h:7959
virtual ~IAlgorithmContext() noexcept=default
Dims getDimensions(int32_t index, OptProfileSelector select) const noexcept
Get the minimum / optimum / maximum dimensions for input or output tensor.
Definition: NvInfer.h:7971
Describes a variation of execution of a layer. An algorithm is represented by IAlgorithmVariant and t...
Definition: NvInfer.h:8009
std::size_t getWorkspaceSize() const noexcept
The size of the GPU temporary memory in bytes which the algorithm uses at execution time.
Definition: NvInfer.h:8030
float getTimingMSec() const noexcept
The time in milliseconds to execute the algorithm.
Definition: NvInfer.h:8022
IAlgorithmIOInfo const * getAlgorithmIOInfoByIndex(int32_t index) const noexcept
Returns the format of an Algorithm input or output. Algorithm inputs are incrementally numbered first...
Definition: NvInfer.h:8044
virtual ~IAlgorithm() noexcept=default
IAlgorithmVariant const & getAlgorithmVariant() const noexcept
Returns the algorithm variant.
Definition: NvInfer.h:8014
Carries information about input or output of the algorithm. IAlgorithmIOInfo for all the input and ou...
Definition: NvInfer.h:7859
virtual ~IAlgorithmIOInfo() noexcept=default
int64_t getVectorizedDim() const noexcept
Return the index of the vectorized dimension or -1 for non-vectorized formats.
Definition: NvInfer.h:7887
Dims getStrides() const noexcept
Return strides of the input/output tensor of algorithm. For vectorized formats, strides are given in ...
Definition: NvInfer.h:7877
DataType getDataType() const noexcept
Return DataType of the input/output of algorithm.
Definition: NvInfer.h:7866
int64_t getComponentsPerElement() const noexcept
Return the number of components per element. This is always 1 for non-vectorized formats.
Definition: NvInfer.h:7898
provides a unique 128-bit identifier, which along with the input and output information denotes the v...
Definition: NvInfer.h:7920
virtual ~IAlgorithmVariant() noexcept=default
int64_t getTactic() const noexcept
Return tactic of the algorithm.
Definition: NvInfer.h:7933
int64_t getImplementation() const noexcept
Return implementation of the algorithm.
Definition: NvInfer.h:7925
An assertion layer in a network.
Definition: NvInfer.h:4877
void setMessage(char const *message) noexcept
Set the message to print if the assertion fails.
Definition: NvInfer.h:4887
char const * getMessage() const noexcept
Return the assertion message.
Definition: NvInfer.h:4897
virtual ~IAssertionLayer() noexcept=default
Holds properties for configuring a builder to produce an engine.
Definition: NvInfer.h:8618
void setMemoryPoolLimit(MemoryPoolType pool, std::size_t poolSize) noexcept
Set the memory size for the memory pool.
Definition: NvInfer.h:9182
void setQuantizationFlag(QuantizationFlag flag) noexcept
Set a single quantization flag.
Definition: NvInfer.h:9049
nvinfer1::ITimingCache * createTimingCache(void const *blob, std::size_t size) const noexcept
Create timing cache.
Definition: NvInfer.h:9117
void setPreviewFeature(PreviewFeature feature, bool enable) noexcept
Enable or disable a specific preview feature.
Definition: NvInfer.h:9219
TRT_DEPRECATED void setInt8Calibrator(IInt8Calibrator *calibrator) noexcept
Set Int8 Calibration interface.
Definition: NvInfer.h:8679
bool getPreviewFeature(PreviewFeature feature) const noexcept
Get status of preview feature.
Definition: NvInfer.h:9233
void clearQuantizationFlag(QuantizationFlag flag) noexcept
clear a quantization flag.
Definition: NvInfer.h:9037
int32_t getBuilderOptimizationLevel() noexcept
Get builder optimization level.
Definition: NvInfer.h:9278
bool setTacticSources(TacticSources tacticSources) noexcept
Set tactic sources.
Definition: NvInfer.h:9083
void setPluginsToSerialize(char const *const *paths, int32_t nbPaths) noexcept
Set the plugin libraries to be serialized with version-compatible engines.
Definition: NvInfer.h:9321
bool getQuantizationFlag(QuantizationFlag flag) const noexcept
Returns true if the quantization flag is set.
Definition: NvInfer.h:9061
TRT_DEPRECATED IInt8Calibrator * getInt8Calibrator() const noexcept
Get Int8 Calibration interface.
Definition: NvInfer.h:8689
std::size_t getMemoryPoolLimit(MemoryPoolType pool) const noexcept
Get the memory size limit of the memory pool.
Definition: NvInfer.h:9201
int32_t getDLACore() const noexcept
Get the DLA core that the engine executes on.
Definition: NvInfer.h:8839
int32_t getNbPluginsToSerialize() const noexcept
Get the number of plugin library paths to be serialized with version-compatible engines.
Definition: NvInfer.h:9344
void setDeviceType(ILayer const *layer, DeviceType deviceType) noexcept
Set the device that this layer must execute on.
Definition: NvInfer.h:8771
void setEngineCapability(EngineCapability capability) noexcept
Configure the builder to target specified EngineCapability flow.
Definition: NvInfer.h:8655
int32_t getMaxAuxStreams() const noexcept
Get the maximum number of auxiliary streams that TRT is allowed to use.
Definition: NvInfer.h:9383
bool getFlag(BuilderFlag builderFlag) const noexcept
Returns true if the build mode flag is set.
Definition: NvInfer.h:8754
void setQuantizationFlags(QuantizationFlags flags) noexcept
Set the quantization flags.
Definition: NvInfer.h:9013
void setProgressMonitor(IProgressMonitor *monitor) noexcept
Sets the progress monitor for building a network.
Definition: NvInfer.h:9399
void setProfilingVerbosity(ProfilingVerbosity verbosity) noexcept
Set verbosity level of layer information exposed in NVTX annotations and IEngineInspector.
Definition: NvInfer.h:8936
IAlgorithmSelector * getAlgorithmSelector() const noexcept
Get Algorithm Selector.
Definition: NvInfer.h:8966
int32_t getNbOptimizationProfiles() const noexcept
Get number of optimization profiles.
Definition: NvInfer.h:8924
QuantizationFlags getQuantizationFlags() const noexcept
Get the quantization flags.
Definition: NvInfer.h:9025
nvinfer1::ITimingCache const * getTimingCache() const noexcept
Get the pointer to the timing cache from current IBuilderConfig.
Definition: NvInfer.h:9150
void reset() noexcept
Resets the builder configuration to defaults.
Definition: NvInfer.h:8870
bool setTimingCache(ITimingCache const &cache, bool ignoreMismatch) noexcept
Attach a timing cache to IBuilderConfig.
Definition: NvInfer.h:9140
char const * getPluginToSerialize(int32_t index) const noexcept
Get the plugin library path to be serialized with version-compatible engines.
Definition: NvInfer.h:9334
EngineCapability getEngineCapability() const noexcept
Query EngineCapability flow configured for the builder.
Definition: NvInfer.h:8667
void setAlgorithmSelector(IAlgorithmSelector *selector) noexcept
Set Algorithm Selector.
Definition: NvInfer.h:8958
DeviceType getDefaultDeviceType() const noexcept
Get the default DeviceType which was set by setDefaultDeviceType.
Definition: NvInfer.h:8860
BuilderFlags getFlags() const noexcept
Get the build mode flags for this builder config. Defaults to 0.
Definition: NvInfer.h:8718
void setFlags(BuilderFlags builderFlags) noexcept
Set the build mode flags to turn on builder options for this network.
Definition: NvInfer.h:8706
TacticSources getTacticSources() const noexcept
Get tactic sources.
Definition: NvInfer.h:9098
void resetDeviceType(ILayer const *layer) noexcept
reset the DeviceType for this layer
Definition: NvInfer.h:8803
void setDLACore(int32_t dlaCore) noexcept
Sets the DLA core used by the network. Defaults to -1.
Definition: NvInfer.h:8829
HardwareCompatibilityLevel getHardwareCompatibilityLevel() const noexcept
Get the hardware compatibility level.
Definition: NvInfer.h:9308
void clearFlag(BuilderFlag builderFlag) noexcept
clear a single build mode flag.
Definition: NvInfer.h:8730
int32_t addOptimizationProfile(IOptimizationProfile const *profile) noexcept
Add an optimization profile.
Definition: NvInfer.h:8911
IProgressMonitor * getProgressMonitor() const noexcept
Definition: NvInfer.h:9409
apiv::VBuilderConfig * mImpl
Definition: NvInfer.h:9415
TRT_DEPRECATED IOptimizationProfile const * getCalibrationProfile() noexcept
Get the current calibration profile.
Definition: NvInfer.h:8996
int32_t getAvgTimingIterations() const noexcept
Query the number of averaging iterations.
Definition: NvInfer.h:8642
void setDefaultDeviceType(DeviceType deviceType) noexcept
Sets the default DeviceType to be used by the builder. It ensures that all the layers that can run on...
Definition: NvInfer.h:8850
void setFlag(BuilderFlag builderFlag) noexcept
Set a single build mode flag.
Definition: NvInfer.h:8742
TRT_DEPRECATED bool setCalibrationProfile(IOptimizationProfile const *profile) noexcept
Add a calibration profile.
Definition: NvInfer.h:8984
virtual ~IBuilderConfig() noexcept=default
DeviceType getDeviceType(ILayer const *layer) const noexcept
Get the device that this layer executes on.
Definition: NvInfer.h:8781
bool canRunOnDLA(ILayer const *layer) const noexcept
Checks if a layer can run on DLA.
Definition: NvInfer.h:8813
cudaStream_t getProfileStream() const noexcept
Get the cuda stream that is used to profile this network.
Definition: NvInfer.h:8894
void setHardwareCompatibilityLevel(HardwareCompatibilityLevel hardwareCompatibilityLevel) noexcept
Set the hardware compatibility level.
Definition: NvInfer.h:9295
void setMaxAuxStreams(int32_t nbStreams) noexcept
Set the maximum number of auxiliary streams that TRT is allowed to use.
Definition: NvInfer.h:9373
ProfilingVerbosity getProfilingVerbosity() const noexcept
Get verbosity level of layer information exposed in NVTX annotations and IEngineInspector.
Definition: NvInfer.h:8949
bool isDeviceTypeSet(ILayer const *layer) const noexcept
whether the DeviceType has been explicitly set for this layer
Definition: NvInfer.h:8793
void setBuilderOptimizationLevel(int32_t level) noexcept
Set builder optimization level.
Definition: NvInfer.h:9266
void setProfileStream(const cudaStream_t stream) noexcept
Set the cuda stream that is used to profile this network.
Definition: NvInfer.h:8882
Builds an engine from a network definition.
Definition: NvInfer.h:9468
int32_t getMaxDLABatchSize() const noexcept
Get the maximum batch size DLA can support. For any tensor the total volume of index dimensions combi...
Definition: NvInfer.h:9495
int32_t getNbDLACores() const noexcept
Return the number of DLA engines available to this builder.
Definition: NvInfer.h:9503
IErrorRecorder * getErrorRecorder() const noexcept
get the ErrorRecorder assigned to this interface.
Definition: NvInfer.h:9601
apiv::VBuilder * mImpl
Definition: NvInfer.h:9714
ILogger * getLogger() const noexcept
get the logger with which the builder was created
Definition: NvInfer.h:9668
bool isNetworkSupported(INetworkDefinition const &network, IBuilderConfig const &config) const noexcept
Checks that a network is within the scope of the IBuilderConfig settings.
Definition: NvInfer.h:9658
bool platformHasTf32() const noexcept
Determine whether the platform has TF32 support.
Definition: NvInfer.h:9617
int32_t getMaxThreads() const noexcept
get the maximum number of threads that can be used by the builder.
Definition: NvInfer.h:9698
IPluginRegistry & getPluginRegistry() noexcept
get the local plugin registry that can be used by the builder.
Definition: NvInfer.h:9708
nvinfer1::IOptimizationProfile * createOptimizationProfile() noexcept
Create a new optimization profile.
Definition: NvInfer.h:9567
void setGpuAllocator(IGpuAllocator *allocator) noexcept
Set the GPU allocator.
Definition: NvInfer.h:9520
nvinfer1::INetworkDefinition * createNetworkV2(NetworkDefinitionCreationFlags flags) noexcept
Create a network definition object.
Definition: NvInfer.h:9552
nvinfer1::IBuilderConfig * createBuilderConfig() noexcept
Create a builder configuration object.
Definition: NvInfer.h:9530
void reset() noexcept
Resets the builder state to default values.
Definition: NvInfer.h:9609
bool setMaxThreads(int32_t maxThreads) noexcept
Set the maximum number of threads.
Definition: NvInfer.h:9684
void setErrorRecorder(IErrorRecorder *recorder) noexcept
Set the ErrorRecorder for this interface.
Definition: NvInfer.h:9586
bool platformHasFastInt8() const noexcept
Determine whether the platform has fast native int8.
Definition: NvInfer.h:9483
nvinfer1::IHostMemory * buildSerializedNetwork(INetworkDefinition &network, IBuilderConfig &config) noexcept
Builds and serializes a network for the given INetworkDefinition and IBuilderConfig.
Definition: NvInfer.h:9636
virtual ~IBuilder() noexcept=default
A cast layer in a network.
Definition: NvInfer.h:3748
virtual ~ICastLayer() noexcept=default
apiv::VCastLayer * mImpl
Definition: NvInfer.h:3774
DataType getToType() const noexcept
Return cast layer output type.
Definition: NvInfer.h:3768
void setToType(DataType toType) noexcept
Set cast layer output type.
Definition: NvInfer.h:3757
A concatenation layer in a network definition.
Definition: NvInfer.h:2052
void setAxis(int32_t axis) noexcept
Set the axis along which concatenation occurs.
Definition: NvInfer.h:2065
int32_t getAxis() const noexcept
Get the axis along which concatenation occurs.
Definition: NvInfer.h:2075
virtual ~IConcatenationLayer() noexcept=default
This layer represents a condition input to an IIfConditional.
Definition: NvInfer.h:4407
virtual ~IConditionLayer() noexcept=default
Layer that represents a constant value.
Definition: NvInfer.h:3787
void setWeights(Weights weights) noexcept
Set the weights for the layer.
Definition: NvInfer.h:3797
Weights getWeights() const noexcept
Get the weights for the layer.
Definition: NvInfer.h:3807
void setDimensions(Dims const &dimensions) noexcept
Set the dimensions for the layer.
Definition: NvInfer.h:3819
apiv::VConstantLayer * mImpl
Definition: NvInfer.h:3837
virtual ~IConstantLayer() noexcept=default
Dims getDimensions() const noexcept
Get the dimensions for the layer.
Definition: NvInfer.h:3831
A convolution layer in a network definition.
Definition: NvInfer.h:1022
Dims getPrePadding() const noexcept
Get the pre-padding.
Definition: NvInfer.h:1147
Weights getBiasWeights() const noexcept
Get the bias weights for the convolution.
Definition: NvInfer.h:1120
void setPaddingMode(PaddingMode paddingMode) noexcept
Set the padding mode.
Definition: NvInfer.h:1188
void setDilationNd(Dims const &dilation) noexcept
Set the multi-dimension dilation of the convolution.
Definition: NvInfer.h:1292
Dims getPaddingNd() const noexcept
Get the multi-dimension padding of the convolution.
Definition: NvInfer.h:1278
Dims getStrideNd() const noexcept
Get the multi-dimension stride of the convolution.
Definition: NvInfer.h:1248
Weights getKernelWeights() const noexcept
Get the kernel weights of the convolution.
Definition: NvInfer.h:1095
void setStrideNd(Dims const &stride) noexcept
Set the multi-dimension stride of the convolution.
Definition: NvInfer.h:1238
Dims getDilationNd() const noexcept
Get the multi-dimension dilation of the convolution.
Definition: NvInfer.h:1302
int64_t getNbOutputMaps() const noexcept
Get the number of output maps for the convolution.
Definition: NvInfer.h:1041
void setKernelWeights(Weights weights) noexcept
Set the kernel weights for the convolution.
Definition: NvInfer.h:1085
Dims getPostPadding() const noexcept
Get the post-padding.
Definition: NvInfer.h:1174
int64_t getNbGroups() const noexcept
Get the number of groups of the convolution.
Definition: NvInfer.h:1071
PaddingMode getPaddingMode() const noexcept
Get the padding mode.
Definition: NvInfer.h:1200
virtual ~IConvolutionLayer() noexcept=default
void setNbGroups(int64_t nbGroups) noexcept
Set the number of groups for a convolution.
Definition: NvInfer.h:1061
void setNbOutputMaps(int64_t nbOutputMaps) noexcept
Set the number of output maps for the convolution.
Definition: NvInfer.h:1031
void setBiasWeights(Weights weights) noexcept
Set the bias weights for the convolution.
Definition: NvInfer.h:1110
Dims getKernelSizeNd() const noexcept
Get the multi-dimension kernel size of the convolution.
Definition: NvInfer.h:1223
void setPaddingNd(Dims const &padding) noexcept
Set the multi-dimension padding of the convolution.
Definition: NvInfer.h:1266
void setPrePadding(Dims const &padding) noexcept
Set the multi-dimension pre-padding of the convolution.
Definition: NvInfer.h:1137
void setPostPadding(Dims const &padding) noexcept
Set the multi-dimension post-padding of the convolution.
Definition: NvInfer.h:1164
void setKernelSizeNd(Dims const &kernelSize) noexcept
Set the multi-dimension kernel size of the convolution.
Definition: NvInfer.h:1213
A deconvolution layer in a network definition.
Definition: NvInfer.h:2093
void setBiasWeights(Weights weights) noexcept
Set the bias weights for the deconvolution.
Definition: NvInfer.h:2181
int64_t getNbGroups() const noexcept
Get the number of groups for a deconvolution.
Definition: NvInfer.h:2142
Weights getKernelWeights() const noexcept
Get the kernel weights for the deconvolution.
Definition: NvInfer.h:2166
void setPrePadding(Dims const &padding) noexcept
Set the multi-dimension pre-padding of the deconvolution.
Definition: NvInfer.h:2209
Dims getStrideNd() const noexcept
Get the multi-dimension stride of the deconvolution.
Definition: NvInfer.h:2325
Dims getDilationNd() const noexcept
Get the multi-dimension dilation of the deconvolution.
Definition: NvInfer.h:2391
Weights getBiasWeights() const noexcept
Get the bias weights for the deconvolution.
Definition: NvInfer.h:2191
void setKernelWeights(Weights weights) noexcept
Set the kernel weights for the deconvolution.
Definition: NvInfer.h:2156
int64_t getNbOutputMaps() const noexcept
Get the number of output feature maps for the deconvolution.
Definition: NvInfer.h:2112
void setStrideNd(Dims const &stride) noexcept
Set the multi-dimension stride of the deconvolution.
Definition: NvInfer.h:2315
Dims getPostPadding() const noexcept
Get the padding.
Definition: NvInfer.h:2247
Dims getKernelSizeNd() const noexcept
Get the multi-dimension kernel size of the deconvolution.
Definition: NvInfer.h:2298
void setPostPadding(Dims const &padding) noexcept
Set the multi-dimension post-padding of the deconvolution.
Definition: NvInfer.h:2237
void setKernelSizeNd(Dims const &kernelSize) noexcept
Set the multi-dimension kernel size of the deconvolution.
Definition: NvInfer.h:2288
virtual ~IDeconvolutionLayer() noexcept=default
void setPaddingNd(Dims const &padding) noexcept
Set the multi-dimension padding of the deconvolution.
Definition: NvInfer.h:2343
void setNbOutputMaps(int64_t nbOutputMaps) noexcept
Set the number of output feature maps for the deconvolution.
Definition: NvInfer.h:2102
Dims getPaddingNd() const noexcept
Get the multi-dimension padding of the deconvolution.
Definition: NvInfer.h:2355
void setDilationNd(Dims const &dilation) noexcept
Set the multi-dimension dilation of the deconvolution.
Definition: NvInfer.h:2381
void setPaddingMode(PaddingMode paddingMode) noexcept
Set the padding mode.
Definition: NvInfer.h:2261
void setNbGroups(int64_t nbGroups) noexcept
Set the number of groups for a deconvolution.
Definition: NvInfer.h:2132
Dims getPrePadding() const noexcept
Get the pre-padding.
Definition: NvInfer.h:2219
PaddingMode getPaddingMode() const noexcept
Get the padding mode.
Definition: NvInfer.h:2273
A Dequantize layer in a network definition.
Definition: NvInfer.h:5463
void setToType(DataType toType) noexcept
Set the Dequantize layer output type.
Definition: NvInfer.h:5500
virtual ~IDequantizeLayer() noexcept=default
int32_t getAxis() const noexcept
Get the quantization axis.
Definition: NvInfer.h:5473
DataType getToType() const noexcept
Return the Dequantize layer output type.
Definition: NvInfer.h:5512
void setAxis(int32_t axis) noexcept
Set the quantization axis.
Definition: NvInfer.h:5484
An Einsum layer in a network.
Definition: NvInfer.h:5559
bool setEquation(char const *equation) noexcept
Set the equation. The equation is a comma-separated list of subscript labels, where each label refers...
Definition: NvInfer.h:5570
virtual ~IEinsumLayer() noexcept=default
char const * getEquation() const noexcept
Return the equation.
Definition: NvInfer.h:5580
A elementwise layer in a network definition.
Definition: NvInfer.h:2467
virtual ~IElementWiseLayer() noexcept=default
apiv::VElementWiseLayer * mImpl
Definition: NvInfer.h:2496
ElementWiseOperation getOperation() const noexcept
Get the binary operation for the layer.
Definition: NvInfer.h:2490
void setOperation(ElementWiseOperation op) noexcept
Set the binary operation for the layer.
Definition: NvInfer.h:2478
Generate a tensor according to a specified mode.
Definition: NvInfer.h:4988
bool isAlphaBetaInt64() const noexcept
Return true if alpha/beta have type int64, false if they have type double.
Definition: NvInfer.h:5220
FillOperation getOperation() const noexcept
Get the fill operation for the layer.
Definition: NvInfer.h:5034
void setOperation(FillOperation op) noexcept
Set the fill operation for the layer.
Definition: NvInfer.h:5024
DataType getToType() const noexcept
Get the fill layer output type.
Definition: NvInfer.h:5249
void setAlphaInt64(int64_t alpha) noexcept
Set the alpha parameter with int64 datatype.
Definition: NvInfer.h:5163
void setBetaInt64(int64_t beta) noexcept
Set the beta parameter with int64 datatype.
Definition: NvInfer.h:5197
void setBeta(double beta) noexcept
Set the beta parameter.
Definition: NvInfer.h:5087
int64_t getAlphaInt64() const noexcept
Get the value of alpha parameter with int64 datatype.
Definition: NvInfer.h:5178
int64_t getBetaInt64() const noexcept
Get the value of beta parameter with int64 datatype.
Definition: NvInfer.h:5212
double getAlpha() const noexcept
Get the value of alpha parameter.
Definition: NvInfer.h:5068
void setDimensions(Dims const &dimensions) noexcept
Set the output tensor's dimensions.
Definition: NvInfer.h:4999
void setAlpha(double alpha) noexcept
Set the alpha parameter.
Definition: NvInfer.h:5053
void setToType(DataType toType) noexcept
Set the fill layer output type.
Definition: NvInfer.h:5237
Dims getDimensions() const noexcept
Get the output tensor's dimensions.
Definition: NvInfer.h:5014
double getBeta() const noexcept
Get the value of beta parameter.
Definition: NvInfer.h:5102
virtual ~IFillLayer() noexcept=default
A Gather layer in a network definition. Supports several kinds of gathering.
Definition: NvInfer.h:2601
void setGatherAxis(int32_t axis) noexcept
Set the axis used by GatherMode::kELEMENTS and GatherMode::kDEFAULT The axis must be less than the nu...
Definition: NvInfer.h:2612
void setNbElementWiseDims(int32_t elementWiseDims) noexcept
Set the number of leading dimensions of indices tensor to be handled elementwise.
Definition: NvInfer.h:2647
apiv::VGatherLayer * mImpl
Definition: NvInfer.h:2683
int32_t getNbElementWiseDims() const noexcept
Get the number of leading dimensions of indices tensor to be handled elementwise.
Definition: NvInfer.h:2657
void setMode(GatherMode mode) noexcept
Set the gather mode.
Definition: NvInfer.h:2667
int32_t getGatherAxis() const noexcept
Get the axis to gather on.
Definition: NvInfer.h:2624
GatherMode getMode() const noexcept
Get the gather mode.
Definition: NvInfer.h:2677
virtual ~IGatherLayer() noexcept=default
A GridSample layer in a network definition.
Definition: NvInfer.h:5779
void setInterpolationMode(InterpolationMode mode) noexcept
Set the grid sample interpolation mode.
Definition: NvInfer.h:5786
bool setSampleMode(SampleMode mode) noexcept
Set the sample mode.
Definition: NvInfer.h:5832
void setAlignCorners(bool alignCorners) noexcept
Set the align corners mode.
Definition: NvInfer.h:5808
apiv::VGridSampleLayer * mImpl
Definition: NvInfer.h:5850
SampleMode getSampleMode() const noexcept
Get the sample mode.
Definition: NvInfer.h:5844
InterpolationMode getInterpolationMode() const noexcept
Get the grid sample interpolation mode.
Definition: NvInfer.h:5798
bool getAlignCorners() const noexcept
Get the align corners mode.
Definition: NvInfer.h:5820
virtual ~IGridSampleLayer() noexcept=default
Class to handle library allocated memory that is accessible to the user.
Definition: NvInferRuntime.h:139
A layer that represents the identity function.
Definition: NvInfer.h:3735
apiv::VIdentityLayer * mImpl
Definition: NvInfer.h:3737
virtual ~IIdentityLayer() noexcept=default
This is a base class for Conditional boundary layers.
Definition: NvInfer.h:4386
IIfConditional * getConditional() const noexcept
Get a pointer to the IIfConditional associated with this boundary layer.
Definition: NvInfer.h:4391
virtual ~IIfConditionalBoundaryLayer() noexcept=default
Helper for constructing conditionally-executed subgraphs.
Definition: NvInfer.h:4466
IIfConditionalInputLayer * addInput(ITensor &input) noexcept
Add an If-conditional input.
Definition: NvInfer.h:4505
char const * getName() const noexcept
Return the name of the conditional.
Definition: NvInfer.h:4530
virtual ~IIfConditional() noexcept=default
IConditionLayer * setCondition(ITensor &condition) noexcept
Set the condition tensor for this If-Conditional construct.
Definition: NvInfer.h:4477
IIfConditionalOutputLayer * addOutput(ITensor &trueSubgraphOutput, ITensor &falseSubgraphOutput) noexcept
Add an If-conditional output.
Definition: NvInfer.h:4493
void setName(char const *name) noexcept
Set the name of the conditional.
Definition: NvInfer.h:4520
This layer represents an input to an IIfConditional.
Definition: NvInfer.h:4435
virtual ~IIfConditionalInputLayer() noexcept=default
This layer represents an output of an IIfConditional.
Definition: NvInfer.h:4422
virtual ~IIfConditionalOutputLayer() noexcept=default
Application-implemented interface for calibration.
Definition: NvInfer.h:7586
virtual TRT_DEPRECATED int32_t getBatchSize() const noexcept=0
Get the batch size used for calibration batches.
A layer to do iterations.
Definition: NvInfer.h:4695
virtual ~IIteratorLayer() noexcept=default
void setReverse(bool reverse) noexcept
Set iteration order to be reverse.
Definition: NvInfer.h:4722
bool getReverse() const noexcept
Check if the iteration order is reverse.
Definition: NvInfer.h:4732
int32_t getAxis() const noexcept
Get axis being iterated over.
Definition: NvInfer.h:4708
void setAxis(int32_t axis) noexcept
Set axis to iterate over.
Definition: NvInfer.h:4700
A LRN layer in a network definition.
Definition: NvInfer.h:1707
int64_t getWindowSize() const noexcept
Get the LRN window size.
Definition: NvInfer.h:1728
float getAlpha() const noexcept
Get the LRN alpha value.
Definition: NvInfer.h:1750
void setWindowSize(int64_t windowSize) noexcept
Set the LRN window size.
Definition: NvInfer.h:1718
void setK(float k) noexcept
Set the LRN K value.
Definition: NvInfer.h:1784
void setAlpha(float alpha) noexcept
Set the LRN alpha value.
Definition: NvInfer.h:1740
void setBeta(float beta) noexcept
Set the LRN beta value.
Definition: NvInfer.h:1762
virtual ~ILRNLayer() noexcept=default
float getBeta() const noexcept
Get the LRN beta value.
Definition: NvInfer.h:1772
float getK() const noexcept
Get the LRN K value.
Definition: NvInfer.h:1794
Base class for all layer classes in a network definition.
Definition: NvInfer.h:550
bool precisionIsSet() const noexcept
whether the computational precision has been set for this layer
Definition: NvInfer.h:692
void setMetadata(char const *metadata) noexcept
Set the metadata for this layer.
Definition: NvInfer.h:808
void setPrecision(DataType dataType) noexcept
Set the preferred or required computational precision of this layer in a weakly-typed network.
Definition: NvInfer.h:668
void setName(char const *name) noexcept
Set the name of a layer.
Definition: NvInfer.h:571
void resetPrecision() noexcept
reset the computational precision for this layer
Definition: NvInfer.h:702
int32_t getNbInputs() const noexcept
Get the number of inputs of a layer.
Definition: NvInfer.h:589
char const * getMetadata() const noexcept
Get the metadata of the layer.
Definition: NvInfer.h:821
DataType getOutputType(int32_t index) const noexcept
get the output type of this layer
Definition: NvInfer.h:764
DataType getPrecision() const noexcept
get the computational precision of this layer
Definition: NvInfer.h:680
char const * getName() const noexcept
Return the name of a layer.
Definition: NvInfer.h:581
int32_t getNbOutputs() const noexcept
Get the number of outputs of a layer.
Definition: NvInfer.h:610
bool outputTypeIsSet(int32_t index) const noexcept
whether the output type has been set for this layer
Definition: NvInfer.h:778
ITensor * getOutput(int32_t index) const noexcept
Get the layer output corresponding to the given index.
Definition: NvInfer.h:620
void setInput(int32_t index, ITensor &tensor) noexcept
Replace an input of this layer with a specific tensor.
Definition: NvInfer.h:637
void resetOutputType(int32_t index) noexcept
reset the output type for this layer
Definition: NvInfer.h:790
ITensor * getInput(int32_t index) const noexcept
Get the layer input corresponding to the given index.
Definition: NvInfer.h:602
void setOutputType(int32_t index, DataType dataType) noexcept
Set the output type of this layer in a weakly-typed network.
Definition: NvInfer.h:749
LayerType getType() const noexcept
Return the type of a layer.
Definition: NvInfer.h:557
virtual ~ILayer() noexcept=default
Application-implemented logging interface for the builder, refitter and runtime.
Definition: NvInferRuntimeBase.h:683
This is a base class for Loop boundary layers.
Definition: NvInfer.h:4363
virtual ~ILoopBoundaryLayer() noexcept=default
ILoop * getLoop() const noexcept
Get a pointer to ILoop associated with this boundary layer.
Definition: NvInfer.h:4368
Helper for creating a recurrent subgraph.
Definition: NvInfer.h:4752
void setName(char const *name) noexcept
Set the name of the loop.
Definition: NvInfer.h:4822
ITripLimitLayer * addTripLimit(ITensor &tensor, TripLimit limit) noexcept
Add a trip-count limiter, based on the given tensor.
Definition: NvInfer.h:4781
IIteratorLayer * addIterator(ITensor &tensor, int32_t axis=0, bool reverse=false) noexcept
Return layer that subscripts tensor by loop iteration.
Definition: NvInfer.h:4794
ILoopOutputLayer * addLoopOutput(ITensor &tensor, LoopOutput outputKind, int32_t axis=0) noexcept
Make an output for this loop, based on the given tensor.
Definition: NvInfer.h:4807
virtual ~ILoop() noexcept=default
char const * getName() const noexcept
Return the name of the loop.
Definition: NvInfer.h:4832
IRecurrenceLayer * addRecurrence(ITensor &initialValue) noexcept
Create a recurrence layer for this loop with initialValue as its first input.
Definition: NvInfer.h:4760
An ILoopOutputLayer is the sole way to get output from a loop.
Definition: NvInfer.h:4595
virtual ~ILoopOutputLayer() noexcept=default
int32_t getAxis() const noexcept
Get axis being concatenated over.
Definition: NvInfer.h:4625
LoopOutput getLoopOutput() const noexcept
Get which kind a loop output has.
Definition: NvInfer.h:4600
void setAxis(int32_t axis) noexcept
Set where to insert the contenation axis. Ignored if getLoopOutput() is kLAST_VALUE.
Definition: NvInfer.h:4617
Layer that represents a Matrix Multiplication.
Definition: NvInfer.h:3627
apiv::VMatrixMultiplyLayer * mImpl
Definition: NvInfer.h:3655
virtual ~IMatrixMultiplyLayer() noexcept=default
MatrixOperation getOperation(int32_t index) const noexcept
Get the operation for an input tensor.
Definition: NvInfer.h:3649
void setOperation(int32_t index, MatrixOperation op) noexcept
Set the operation for an input tensor.
Definition: NvInfer.h:3637
A non-maximum suppression layer in a network definition.
Definition: NvInfer.h:5927
virtual ~INMSLayer() noexcept=default
void setTopKBoxLimit(int32_t limit) noexcept
Set the TopK box limit parameter for the layer.
Definition: NvInfer.h:5964
void setBoundingBoxFormat(BoundingBoxFormat fmt) noexcept
Set the bounding box format parameter for the layer.
Definition: NvInfer.h:5938
BoundingBoxFormat getBoundingBoxFormat() const noexcept
Get the bounding box format parameter for the layer.
Definition: NvInfer.h:5950
apiv::VNMSLayer * mImpl
Definition: NvInfer.h:6000
int32_t getTopKBoxLimit() const noexcept
Get the TopK box limit parameter for the layer.
Definition: NvInfer.h:5974
A network definition for input to the builder.
Definition: NvInfer.h:6223
IPluginV2Layer * addPluginV2(ITensor *const *inputs, int32_t nbInputs, IPluginV2 &plugin) noexcept
Add a plugin layer to the network using the IPluginV2 interface.
Definition: NvInfer.h:6825
IConcatenationLayer * addConcatenation(ITensor *const *inputs, int32_t nbInputs) noexcept
Add a concatenation layer to the network.
Definition: NvInfer.h:6416
IShuffleLayer * addShuffle(ITensor &input) noexcept
Add a shuffle layer to the network.
Definition: NvInfer.h:6479
INormalizationLayer * addNormalization(ITensor &input, ITensor &scale, ITensor &bias, uint32_t axesMask) noexcept
Add a normalization layer to the network.
Definition: NvInfer.h:7526
void setName(char const *name) noexcept
Sets the name of the network.
Definition: NvInfer.h:6887
bool markDebug(ITensor &tensor) noexcept
Mark a tensor as a debug tensor.
Definition: NvInfer.h:6295
ILRNLayer * addLRN(ITensor &input, int64_t window, float alpha, float beta, float k) noexcept
Add a LRN layer to the network.
Definition: NvInfer.h:6360
ITopKLayer * addTopK(ITensor &input, TopKOperation op, int32_t k, uint32_t reduceAxes) noexcept
Add a TopK layer to the network.
Definition: NvInfer.h:6639
IAssertionLayer * addAssertion(ITensor &condition, char const *message) noexcept
Add an assertion layer to the network.
Definition: NvInfer.h:7203
IConvolutionLayer * addConvolutionNd(ITensor &input, int64_t nbOutputMaps, Dims const &kernelSize, Weights kernelWeights, Weights biasWeights) noexcept
Add a multi-dimension convolution layer to the network.
Definition: NvInfer.h:7022
ICastLayer * addCast(ITensor &input, DataType toType) noexcept
Add a cast layer.
Definition: NvInfer.h:6779
IScaleLayer * addScaleNd(ITensor &input, ScaleMode mode, Weights shift, Weights scale, Weights power, int32_t channelAxis) noexcept
Add a multi-dimension scale layer to the network.
Definition: NvInfer.h:7101
char const * getName() const noexcept
Returns the name associated with the network.
Definition: NvInfer.h:6901
IParametricReLULayer * addParametricReLU(ITensor &input, ITensor &slope) noexcept
Add a parametric ReLU layer to the network.
Definition: NvInfer.h:7000
ITensor * getOutput(int32_t index) const noexcept
Get the output tensor specified by the given index.
Definition: NvInfer.h:6580
ITensor * getInput(int32_t index) const noexcept
Get the input tensor specified by the given index.
Definition: NvInfer.h:6550
IDequantizeLayer * addDequantize(ITensor &input, ITensor &scale, DataType outputType) noexcept
Add a dequantization layer to the network.
Definition: NvInfer.h:7370
bool unmarkOutputForShapes(ITensor &tensor) noexcept
Undo markOutputForShapes.
Definition: NvInfer.h:6982
IFillLayer * addFill(Dims const &dimensions, FillOperation op, DataType outputType) noexcept
Add a fill layer to the network.
Definition: NvInfer.h:7254
ILoop * addLoop() noexcept
Add a loop to the network.
Definition: NvInfer.h:7132
IActivationLayer * addActivation(ITensor &input, ActivationType type) noexcept
Add an activation layer to the network.
Definition: NvInfer.h:6341
TRT_DEPRECATED IFillLayer * addFill(Dims const &dimensions, FillOperation op) noexcept
Add a fill layer to the network.
Definition: NvInfer.h:7228
ISliceLayer * addSlice(ITensor &input, Dims const &start, Dims const &size, Dims const &stride) noexcept
Add a slice layer to the network.
Definition: NvInfer.h:6863
virtual ~INetworkDefinition() noexcept=default
TRT_DEPRECATED IQuantizeLayer * addQuantize(ITensor &input, ITensor &scale) noexcept
Add a quantization layer to the network.
Definition: NvInfer.h:7411
virtual IBuilder & getBuilder() const noexcept
Return the builder from which this INetworkDefinition was created.
Definition: NvInfer.h:7537
INMSLayer * addNMS(ITensor &boxes, ITensor &scores, ITensor &maxOutputBoxesPerClass) noexcept
Add a non-maximum suppression layer to the network.
Definition: NvInfer.h:7483
ILayer * getLayer(int32_t index) const noexcept
Get the layer specified by the given index.
Definition: NvInfer.h:6522
bool getFlag(NetworkDefinitionCreationFlag networkDefinitionCreationFlag) const noexcept
Returns true if the network definition creation flag is set.
Definition: NvInfer.h:6953
IIfConditional * addIfConditional() noexcept
Add an if-then-else to the network.
Definition: NvInfer.h:7147
IErrorRecorder * getErrorRecorder() const noexcept
get the ErrorRecorder assigned to this interface.
Definition: NvInfer.h:7328
IReverseSequenceLayer * addReverseSequence(ITensor &input, ITensor &sequenceLens) noexcept
Add a ReverseSequence layer to the network.
Definition: NvInfer.h:7500
int32_t getNbInputs() const noexcept
Get the number of inputs in the network.
Definition: NvInfer.h:6534
NetworkDefinitionCreationFlags getFlags() const noexcept
Get the network definition creation flags for this network definition object. Defaults to 0.
Definition: NvInfer.h:6941
IQuantizeLayer * addQuantize(ITensor &input, ITensor &scale, DataType outputType) noexcept
Add a quantization layer to the network.
Definition: NvInfer.h:7432
IReduceLayer * addReduce(ITensor &input, ReduceOperation operation, uint32_t reduceAxes, bool keepDimensions) noexcept
Add a reduce layer to the network.
Definition: NvInfer.h:6606
IUnaryLayer * addUnary(ITensor &input, UnaryOperation operation) noexcept
Add a unary layer to the network.
Definition: NvInfer.h:6465
IGridSampleLayer * addGridSample(ITensor &input, ITensor &grid) noexcept
Add a GridSample layer to the network.
Definition: NvInfer.h:7465
void removeTensor(ITensor &tensor) noexcept
remove a tensor from the network definition.
Definition: NvInfer.h:6794
ISelectLayer * addSelect(ITensor &condition, ITensor &thenInput, ITensor &elseInput) noexcept
Add a select layer to the network.
Definition: NvInfer.h:7186
IScatterLayer * addScatter(ITensor &data, ITensor &indices, ITensor &updates, ScatterMode mode) noexcept
Add a Scatter layer to the network with specified mode and axis=0.
Definition: NvInfer.h:7390
int32_t getNbLayers() const noexcept
Get the number of layers in the network.
Definition: NvInfer.h:6508
TRT_DEPRECATED bool hasImplicitBatchDimension() const noexcept
Query whether the network was created with an implicit batch dimension.
Definition: NvInfer.h:6931
apiv::VNetworkDefinition * mImpl
Definition: NvInfer.h:7543
bool markOutputForShapes(ITensor &tensor) noexcept
Enable tensor's value to be computed by IExecutionContext::getShapeBinding.
Definition: NvInfer.h:6970
IOneHotLayer * addOneHot(ITensor &indices, ITensor &values, ITensor &depth, int32_t axis) noexcept
Add a OneHot layer to the network.
Definition: NvInfer.h:6496
IScaleLayer * addScale(ITensor &input, ScaleMode mode, Weights shift, Weights scale, Weights power) noexcept
Add a Scale layer to the network.
Definition: NvInfer.h:6386
IPluginV3Layer * addPluginV3(ITensor *const *inputs, int32_t nbInputs, ITensor *const *shapeInputs, int32_t nbShapeInputs, IPluginV3 &plugin) noexcept
Add a plugin layer implementing the IPluginV3 interface to the network.
Definition: NvInfer.h:6843
void unmarkOutput(ITensor &tensor) noexcept
unmark a tensor as a network output.
Definition: NvInfer.h:6806
IIdentityLayer * addIdentity(ITensor &input) noexcept
Add an identity layer.
Definition: NvInfer.h:6764
IGatherLayer * addGatherV2(ITensor &data, ITensor &indices, GatherMode mode) noexcept
Add gather with specified mode, axis=0 and nbElementWiseDims=0.
Definition: NvInfer.h:6671
IElementWiseLayer * addElementWise(ITensor &input1, ITensor &input2, ElementWiseOperation op) noexcept
Add an elementwise layer to the network.
Definition: NvInfer.h:6443
IConstantLayer * addConstant(Dims const &dimensions, Weights weights) noexcept
Add a constant layer to the network.
Definition: NvInfer.h:6750
void setErrorRecorder(IErrorRecorder *recorder) noexcept
Set the ErrorRecorder for this interface.
Definition: NvInfer.h:7313
IPoolingLayer * addPoolingNd(ITensor &input, PoolingType type, Dims const &windowSize) noexcept
Add a multi-dimension pooling layer to the network.
Definition: NvInfer.h:7042
IRaggedSoftMaxLayer * addRaggedSoftMax(ITensor &input, ITensor &bounds) noexcept
Add a RaggedSoftMax layer to the network.
Definition: NvInfer.h:6690
IShapeLayer * addShape(ITensor &input) noexcept
Add a shape layer to the network.
Definition: NvInfer.h:6917
IGatherLayer * addGather(ITensor &data, ITensor &indices, int32_t axis) noexcept
Add gather with mode GatherMode::kDEFAULT and specified axis and nbElementWiseDims=0.
Definition: NvInfer.h:6655
IDeconvolutionLayer * addDeconvolutionNd(ITensor &input, int64_t nbOutputMaps, Dims kernelSize, Weights kernelWeights, Weights biasWeights) noexcept
Add a multi-dimension deconvolution layer to the network.
Definition: NvInfer.h:7064
IResizeLayer * addResize(ITensor &input) noexcept
Add a resize layer to the network.
Definition: NvInfer.h:7118
IMatrixMultiplyLayer * addMatrixMultiply(ITensor &input0, MatrixOperation op0, ITensor &input1, MatrixOperation op1) noexcept
Add a MatrixMultiply layer to the network.
Definition: NvInfer.h:6711
ISoftMaxLayer * addSoftMax(ITensor &input) noexcept
Add a SoftMax layer to the network.
Definition: NvInfer.h:6399
bool isDebugTensor(nvinfer1::ITensor const &tensor) const noexcept
Check if a tensor is marked as debug tensor.
Definition: NvInfer.h:6321
bool unmarkDebug(ITensor &tensor) noexcept
Unmark a tensor as a debug tensor.
Definition: NvInfer.h:6311
IEinsumLayer * addEinsum(ITensor *const *inputs, int32_t nbInputs, char const *equation) noexcept
Add an Einsum layer to the network.
Definition: NvInfer.h:7447
void markOutput(ITensor &tensor) noexcept
Mark a tensor as a network output.
Definition: NvInfer.h:6277
IPaddingLayer * addPaddingNd(ITensor &input, Dims const &prePadding, Dims const &postPadding) noexcept
Add a padding layer to the network. Only 2D padding is currently supported.
Definition: NvInfer.h:7270
INonZeroLayer * addNonZero(ITensor &input) noexcept
Add a nonzero layer to the network.
Definition: NvInfer.h:6726
TRT_DEPRECATED IDequantizeLayer * addDequantize(ITensor &input, ITensor &scale) noexcept
Add a dequantization layer to the network.
Definition: NvInfer.h:7349
int32_t getNbOutputs() const noexcept
Get the number of outputs in the network.
Definition: NvInfer.h:6564
bool setWeightsName(Weights weights, char const *name) noexcept
Associate a name with all current uses of the given weights.
Definition: NvInfer.h:7294
Forward declaration of IEngineInspector for use by other interfaces.
Definition: NvInferRuntime.h:48
Definition: NvInfer.h:3681
virtual ~INonZeroLayer() noexcept=default
A normalization layer in a network definition.
Definition: NvInfer.h:6089
float getEpsilon() const noexcept
Get the epsilon value used for the normalization calculation.
Definition: NvInfer.h:6108
uint32_t getAxes() const noexcept
Get the axes value used for the normalization calculation.
Definition: NvInfer.h:6128
virtual ~INormalizationLayer() noexcept=default
void setEpsilon(float eps) noexcept
Set the epsilon value used for the normalization calculation.
Definition: NvInfer.h:6098
DataType getComputePrecision() const noexcept
Get the compute precision of this layer.
Definition: NvInfer.h:6195
apiv::VNormalizationLayer * mImpl
Definition: NvInfer.h:6201
int64_t getNbGroups() const noexcept
Get the number of groups used to split the channels for the normalization calculation.
Definition: NvInfer.h:6159
void setAxes(uint32_t axesMask) noexcept
Set the reduction axes for the normalization calculation.
Definition: NvInfer.h:6118
void setComputePrecision(DataType type) noexcept
Set the compute precision of this layer.
Definition: NvInfer.h:6185
void setNbGroups(int64_t nbGroups) noexcept
Set the number of groups used to split the channels in the normalization calculation.
Definition: NvInfer.h:6149
A OneHot layer in a network definition.
Definition: NvInfer.h:5743
apiv::VOneHotLayer * mImpl
Definition: NvInfer.h:5764
void setAxis(int32_t axis) noexcept
Set the axis parameter.
Definition: NvInfer.h:5750
int32_t getAxis() const noexcept
Get the value of the axis parameter.
Definition: NvInfer.h:5758
Optimization profile for dynamic input dimensions and shape tensors.
Definition: NvInferRuntime.h:2104
Layer that represents a padding operation.
Definition: NvInfer.h:2958
Dims getPostPaddingNd() const noexcept
Get the padding that is applied at the end of the tensor.
Definition: NvInfer.h:3007
void setPrePaddingNd(Dims const &padding) noexcept
Set the padding that is applied at the start of the tensor.
Definition: NvInfer.h:2969
virtual ~IPaddingLayer() noexcept=default
void setPostPaddingNd(Dims const &padding) noexcept
Set the padding that is applied at the end of the tensor.
Definition: NvInfer.h:2995
Dims getPrePaddingNd() const noexcept
Get the padding that is applied at the start of the tensor.
Definition: NvInfer.h:2981
apiv::VPaddingLayer * mImpl
Definition: NvInfer.h:3013
Layer that represents a parametric ReLU operation.
Definition: NvInfer.h:3851
apiv::VParametricReLULayer * mImpl
Definition: NvInfer.h:3853
virtual ~IParametricReLULayer() noexcept=default
Single registration point for all plugins in an application. It is used to find plugin implementation...
Definition: NvInferRuntimeCommon.h:56
Plugin class for user-implemented layers.
Definition: NvInferRuntimePlugin.h:128
Layer type for pluginV2.
Definition: NvInfer.h:2697
virtual ~IPluginV2Layer() noexcept=default
apiv::VPluginV2Layer * mImpl
Definition: NvInfer.h:2710
IPluginV2 & getPlugin() noexcept
Get the plugin for the layer.
Definition: NvInfer.h:2704
Layer type for V3 plugins.
Definition: NvInfer.h:2724
virtual ~IPluginV3Layer() noexcept=default
IPluginV3 & getPlugin() noexcept
Get the plugin for the layer.
Definition: NvInfer.h:2731
apiv::VPluginV3Layer * mImpl
Definition: NvInfer.h:2737
A Pooling layer in a network definition.
Definition: NvInfer.h:1456
PoolingType getPoolingType() const noexcept
Get the type of activation to be performed.
Definition: NvInfer.h:1475
PaddingMode getPaddingMode() const noexcept
Get the padding mode.
Definition: NvInfer.h:1608
Dims getPostPadding() const noexcept
Get the padding.
Definition: NvInfer.h:1584
bool getAverageCountExcludesPadding() const noexcept
Get whether average pooling uses as a denominator the overlap area between the window and the unpadde...
Definition: NvInfer.h:1528
Dims getPrePadding() const noexcept
Get the pre-padding.
Definition: NvInfer.h:1556
void setPoolingType(PoolingType type) noexcept
Set the type of activation to be performed.
Definition: NvInfer.h:1465
void setWindowSizeNd(Dims const &windowSize) noexcept
Set the multi-dimension window size for pooling.
Definition: NvInfer.h:1621
void setPaddingMode(PaddingMode paddingMode) noexcept
Set the padding mode.
Definition: NvInfer.h:1597
Dims getWindowSizeNd() const noexcept
Get the multi-dimension window size for pooling.
Definition: NvInfer.h:1631
void setAverageCountExcludesPadding(bool exclusive) noexcept
Set whether average pooling uses as a denominator the overlap area between the window and the unpadde...
Definition: NvInfer.h:1517
void setPaddingNd(Dims const &padding) noexcept
Set the multi-dimension padding for pooling.
Definition: NvInfer.h:1675
float getBlendFactor() const noexcept
Get the blending factor for the max_average_blend mode: max_average_blendPool = (1-blendFactor)*maxPo...
Definition: NvInfer.h:1503
void setStrideNd(Dims const &stride) noexcept
Set the multi-dimension stride for pooling.
Definition: NvInfer.h:1646
Dims getStrideNd() const noexcept
Get the multi-dimension stride for pooling.
Definition: NvInfer.h:1656
virtual ~IPoolingLayer() noexcept=default
Dims getPaddingNd() const noexcept
Get the multi-dimension padding for pooling.
Definition: NvInfer.h:1687
void setPostPadding(Dims const &padding) noexcept
Set the multi-dimension post-padding for pooling.
Definition: NvInfer.h:1574
void setPrePadding(Dims const &padding) noexcept
Set the multi-dimension pre-padding for pooling.
Definition: NvInfer.h:1546
void setBlendFactor(float blendFactor) noexcept
Set the blending factor for the max_average_blend mode: max_average_blendPool = (1-blendFactor)*maxPo...
Definition: NvInfer.h:1490
A Quantize layer in a network definition.
Definition: NvInfer.h:5333
void setToType(DataType toType) noexcept
Set the Quantize layer output type.
Definition: NvInfer.h:5370
void setAxis(int32_t axis) noexcept
Set the quantization axis.
Definition: NvInfer.h:5354
int32_t getAxis() const noexcept
Get the quantization axis.
Definition: NvInfer.h:5343
virtual ~IQuantizeLayer() noexcept=default
DataType getToType() const noexcept
Return the Quantize layer output type.
Definition: NvInfer.h:5382
A RaggedSoftmax layer in a network definition.
Definition: NvInfer.h:3702
apiv::VRaggedSoftMaxLayer * mImpl
Definition: NvInfer.h:3704
virtual ~IRaggedSoftMaxLayer() noexcept=default
A recurrence layer in a network definition.
Definition: NvInfer.h:4548
virtual ~IRecurrenceLayer() noexcept=default
Layer that represents a reduction across a non-bool tensor.
Definition: NvInfer.h:2880
void setKeepDimensions(bool keepDimensions) noexcept
Set the boolean that specifies whether or not to keep the reduced dimensions for the layer.
Definition: NvInfer.h:2927
void setOperation(ReduceOperation op) noexcept
Set the reduce operation for the layer.
Definition: NvInfer.h:2887
ReduceOperation getOperation() const noexcept
Get the reduce operation for the layer.
Definition: NvInfer.h:2897
virtual ~IReduceLayer() noexcept=default
uint32_t getReduceAxes() const noexcept
Get the axes over which to reduce for the layer.
Definition: NvInfer.h:2917
void setReduceAxes(uint32_t reduceAxes) noexcept
Set the axes over which to reduce.
Definition: NvInfer.h:2907
apiv::VReduceLayer * mImpl
Definition: NvInfer.h:2943
bool getKeepDimensions() const noexcept
Get the boolean that specifies whether or not to keep the reduced dimensions for the layer.
Definition: NvInfer.h:2937
A resize layer in a network definition.
Definition: NvInfer.h:4040
void setSelectorForSinglePixel(ResizeSelector selector) noexcept
Set coordinate selector function when resized to single pixel.
Definition: NvInfer.h:4201
void setNearestRounding(ResizeRoundMode value) noexcept
Set rounding mode for nearest neighbor resize.
Definition: NvInfer.h:4225
virtual ~IResizeLayer() noexcept=default
int32_t getScales(int32_t size, float *scales) const noexcept
Copies resize scales to scales[0, ..., nbScales-1], where nbScales is the number of scales that were ...
Definition: NvInfer.h:4119
void setOutputDimensions(Dims const &dimensions) noexcept
Set the output dimensions.
Definition: NvInfer.h:4060
void setCubicCoeff(float A) noexcept
Set the coefficient 'A' used in cubic interpolation.
Definition: NvInfer.h:4257
void setScales(float const *scales, int32_t nbScales) noexcept
Set the resize scales.
Definition: NvInfer.h:4100
float getCubicCoeff() const noexcept
Get the coefficient 'A' used in cubic interpolation.
Definition: NvInfer.h:4267
ResizeSelector getSelectorForSinglePixel() const noexcept
Get the coordinate selector function when resized to single pixel.
Definition: NvInfer.h:4211
InterpolationMode getResizeMode() const noexcept
Get resize mode for an input tensor.
Definition: NvInfer.h:4141
void setCoordinateTransformation(ResizeCoordinateTransformation coordTransform) noexcept
Set coordinate transformation function.
Definition: NvInfer.h:4176
void setExcludeOutside(bool excludeFlag) noexcept
Set the state for excluding outside pixels.
Definition: NvInfer.h:4280
void setResizeMode(InterpolationMode interpolationMode) noexcept
Set resize mode for an input tensor.
Definition: NvInfer.h:4131
Dims getOutputDimensions() const noexcept
Get the output dimensions.
Definition: NvInfer.h:4070
ResizeRoundMode getNearestRounding() const noexcept
Get rounding mode for nearest neighbor resize.
Definition: NvInfer.h:4235
bool getExcludeOutside() const noexcept
Get the state for excluding outside pixels.
Definition: NvInfer.h:4290
ResizeCoordinateTransformation getCoordinateTransformation() const noexcept
Get coordinate transformation function.
Definition: NvInfer.h:4186
A ReverseSequence layer in a network definition.
Definition: NvInfer.h:6017
void setSequenceAxis(int32_t sequenceAxis) noexcept
Set the sequence axis. Default is 0.
Definition: NvInfer.h:6050
int32_t getBatchAxis() const noexcept
Return the batch axis. Return 1 if no batch axis was set.
Definition: NvInfer.h:6037
apiv::VReverseSequenceLayer * mImpl
Definition: NvInfer.h:6066
int32_t getSequenceAxis() const noexcept
Return the sequence axis. Return 0 if no sequence axis was set.
Definition: NvInfer.h:6060
void setBatchAxis(int32_t batchAxis) noexcept
Set the batch axis. Default is 1.
Definition: NvInfer.h:6027
virtual ~IReverseSequenceLayer() noexcept=default
A Scale layer in a network definition.
Definition: NvInfer.h:1853
Weights getScale() const noexcept
Get the scale value.
Definition: NvInfer.h:1910
Weights getPower() const noexcept
Get the power value.
Definition: NvInfer.h:1930
void setScale(Weights scale) noexcept
Set the scale value.
Definition: NvInfer.h:1900
void setPower(Weights power) noexcept
Set the power value.
Definition: NvInfer.h:1920
ScaleMode getMode() const noexcept
Get the scale mode.
Definition: NvInfer.h:1870
void setShift(Weights shift) noexcept
Set the shift value.
Definition: NvInfer.h:1880
void setChannelAxis(int32_t channelAxis) noexcept
Set the channel axis.
Definition: NvInfer.h:1966
Weights getShift() const noexcept
Get the shift value.
Definition: NvInfer.h:1890
virtual ~IScaleLayer() noexcept=default
void setMode(ScaleMode mode) noexcept
Set the scale mode.
Definition: NvInfer.h:1860
int32_t getChannelAxis() const noexcept
Get the channel axis.
Definition: NvInfer.h:1945
A scatter layer in a network definition. Supports several kinds of scattering.
Definition: NvInfer.h:5671
void setMode(ScatterMode mode) noexcept
Set the scatter mode.
Definition: NvInfer.h:5678
apiv::VScatterLayer * mImpl
Definition: NvInfer.h:5712
void setAxis(int32_t axis) noexcept
Set the axis used by ScatterMode::kELEMENTS.
Definition: NvInfer.h:5698
int32_t getAxis() const noexcept
Get the axis.
Definition: NvInfer.h:5706
ScatterMode getMode() const noexcept
Get the scatter mode.
Definition: NvInfer.h:5688
virtual ~IScatterLayer() noexcept=default
Select elements from two data tensors based on a condition tensor.
Definition: NvInfer.h:4855
virtual ~ISelectLayer() noexcept=default
Layer type for getting shape of a tensor.
Definition: NvInfer.h:3430
virtual ~IShapeLayer() noexcept=default
apiv::VShapeLayer * mImpl
Definition: NvInfer.h:3432
Layer type for shuffling data.
Definition: NvInfer.h:3046
apiv::VShuffleLayer * mImpl
Definition: NvInfer.h:3204
void setFirstTranspose(Permutation permutation) noexcept
Set the permutation applied by the first transpose operation.
Definition: NvInfer.h:3057
void setSecondTranspose(Permutation permutation) noexcept
Set the permutation applied by the second transpose operation.
Definition: NvInfer.h:3157
Dims getReshapeDimensions() const noexcept
Get the reshaped dimensions.
Definition: NvInfer.h:3110
void setReshapeDimensions(Dims const &dimensions) noexcept
Set the reshaped dimensions.
Definition: NvInfer.h:3097
Permutation getFirstTranspose() const noexcept
Get the permutation applied by the first transpose operation.
Definition: NvInfer.h:3069
virtual ~IShuffleLayer() noexcept=default
Permutation getSecondTranspose() const noexcept
Get the permutation applied by the second transpose operation.
Definition: NvInfer.h:3169
bool getZeroIsPlaceholder() const noexcept
Get meaning of 0 in reshape dimensions.
Definition: NvInfer.h:3198
void setZeroIsPlaceholder(bool zeroIsPlaceholder) noexcept
Set meaning of 0 in reshape dimensions.
Definition: NvInfer.h:3185
Slices an input tensor into an output tensor based on the offset and strides.
Definition: NvInfer.h:3278
void setStride(Dims const &stride) noexcept
Set the stride for computing the output slice data.
Definition: NvInfer.h:3347
apiv::VSliceLayer * mImpl
Definition: NvInfer.h:3413
virtual ~ISliceLayer() noexcept=default
void setSize(Dims const &size) noexcept
Set the dimensions of the output slice.
Definition: NvInfer.h:3318
void setStart(Dims const &start) noexcept
Set the start offset that the slice layer uses to create the output slice.
Definition: NvInfer.h:3289
Dims getStart() const noexcept
Get the start offset for the slice layer.
Definition: NvInfer.h:3304
void setMode(SampleMode mode) noexcept
Set the slice mode.
Definition: NvInfer.h:3372
Dims getSize() const noexcept
Get dimensions of the output slice.
Definition: NvInfer.h:3333
SampleMode getMode() const noexcept
Get the slice mode.
Definition: NvInfer.h:3382
Dims getStride() const noexcept
Get the stride for the output slice.
Definition: NvInfer.h:3362
A Softmax layer in a network definition.
Definition: NvInfer.h:1997
void setAxes(uint32_t axes) noexcept
Set the axis along which softmax is computed. Currently, only one axis can be set.
Definition: NvInfer.h:2019
uint32_t getAxes() const noexcept
Get the axis along which softmax occurs.
Definition: NvInfer.h:2029
virtual ~ISoftMaxLayer() noexcept=default
A tensor in a network definition.
Definition: NvInfer.h:181
void setAllowedFormats(TensorFormats formats) noexcept
Set allowed formats for an input or output tensor. By default all formats are allowed....
Definition: NvInfer.h:426
TensorLocation getLocation() const noexcept
Get the storage location of a tensor.
Definition: NvInfer.h:345
void setDimensions(Dims const &dimensions) noexcept
Set the dimensions of a tensor.
Definition: NvInfer.h:228
void resetDynamicRange() noexcept
Undo effect of setDynamicRange.
Definition: NvInfer.h:384
void setName(char const *name) noexcept
Set the tensor name.
Definition: NvInfer.h:197
bool isExecutionTensor() const noexcept
Whether the tensor is an execution tensor.
Definition: NvInfer.h:491
void setType(DataType type) noexcept
Set the data type of a tensor.
Definition: NvInfer.h:257
bool dynamicRangeIsSet() const noexcept
Query whether dynamic range is set.
Definition: NvInfer.h:376
char const * getName() const noexcept
Get the tensor name.
Definition: NvInfer.h:209
bool isShapeTensor() const noexcept
Whether the tensor is a shape tensor.
Definition: NvInfer.h:470
float getDynamicRangeMax() const noexcept
Get maximum of dynamic range.
Definition: NvInfer.h:404
bool isNetworkInput() const noexcept
Whether the tensor is a network input.
Definition: NvInfer.h:294
TRT_DEPRECATED void setBroadcastAcrossBatch(bool broadcastAcrossBatch) noexcept
Set whether to enable broadcast of tensor across the implicit batch dimension.
Definition: NvInfer.h:319
TRT_DEPRECATED bool setDynamicRange(float min, float max) noexcept
Set dynamic range for the tensor.
Definition: NvInfer.h:286
TRT_DEPRECATED bool getBroadcastAcrossBatch() const noexcept
Check if tensor is broadcast across the implicit batch dimension.
Definition: NvInfer.h:333
bool isNetworkOutput() const noexcept
Whether the tensor is a network output.
Definition: NvInfer.h:302
DataType getType() const noexcept
Get the data type of a tensor.
Definition: NvInfer.h:269
apiv::VTensor * mImpl
Definition: NvInfer.h:538
float getDynamicRangeMin() const noexcept
Get minimum of dynamic range.
Definition: NvInfer.h:394
virtual ~ITensor() noexcept=default
void setDimensionName(int32_t index, char const *name) noexcept
Name a dimension of an input tensor.
Definition: NvInfer.h:517
char const * getDimensionName(int32_t index) const noexcept
Get the name of an input dimension.
Definition: NvInfer.h:532
TRT_DEPRECATED void setLocation(TensorLocation location) noexcept
Set the storage location of a tensor.
Definition: NvInfer.h:364
Dims getDimensions() const noexcept
Get the dimensions of a tensor.
Definition: NvInfer.h:242
TensorFormats getAllowedFormats() const noexcept
Get a bitmask of TensorFormat values that the tensor supports. For a shape tensor,...
Definition: NvInfer.h:439
Class to handle tactic timing info collected from builder.
Definition: NvInfer.h:8319
bool combine(ITimingCache const &inputCache, bool ignoreMismatch) noexcept
Combine input timing cache into local instance.
Definition: NvInfer.h:8356
virtual ~ITimingCache() noexcept=default
apiv::VTimingCache * mImpl
Definition: NvInfer.h:8372
bool reset() noexcept
Empty the timing cache.
Definition: NvInfer.h:8366
Layer that represents a TopK reduction.
Definition: NvInfer.h:3470
void setK(int32_t k) noexcept
Set the static k value for the layer.
Definition: NvInfer.h:3501
void setReduceAxes(uint32_t reduceAxes) noexcept
Set which axes to reduce for the layer.
Definition: NvInfer.h:3525
TopKOperation getOperation() const noexcept
Get the operation for the layer.
Definition: NvInfer.h:3487
apiv::VTopKLayer * mImpl
Definition: NvInfer.h:3557
void setOperation(TopKOperation op) noexcept
Set the operation for the layer.
Definition: NvInfer.h:3477
int32_t getK() const noexcept
Get the k value for the layer.
Definition: NvInfer.h:3515
uint32_t getReduceAxes() const noexcept
Get the axes to reduce for the layer.
Definition: NvInfer.h:3535
virtual ~ITopKLayer() noexcept=default
A layer that represents a trip-count limiter.
Definition: NvInfer.h:4669
TripLimit getTripLimit() const noexcept
Get a trip limiter type.
Definition: NvInfer.h:4674
virtual ~ITripLimitLayer() noexcept=default
Layer that represents an unary operation.
Definition: NvInfer.h:2805
void setOperation(UnaryOperation op) noexcept
Set the unary operation for the layer.
Definition: NvInfer.h:2814
apiv::VUnaryLayer * mImpl
Definition: NvInfer.h:2830
UnaryOperation getOperation() const noexcept
Get the unary operation for the layer.
Definition: NvInfer.h:2824
virtual ~IUnaryLayer() noexcept=default
An Interface class for version control.
Definition: NvInferRuntimeBase.h:400
Version information associated with a TRT interface.
Definition: NvInferRuntimeBase.h:365
An array of weights used as a layer parameter.
Definition: NvInferRuntime.h:121
Single registration point for all plugins in an application. It is used to find plugin implementation...
Definition: NvInferSafeRuntime.h:830
Definition: NvInfer.h:8057
virtual int32_t selectAlgorithms(IAlgorithmContext const &context, IAlgorithm const *const *choices, int32_t nbChoices, int32_t *selection) noexcept=0
Select Algorithms for a layer from the given list of algorithm choices.
virtual void reportAlgorithms(IAlgorithmContext const *const *algoContexts, IAlgorithm const *const *algoChoices, int32_t nbAlgorithms) noexcept=0
Called by TensorRT to report choices it made.
InterfaceInfo getInterfaceInfo() const noexcept override
Return version information associated with this interface. Applications must not override this method...
Definition: NvInfer.h:8062
virtual ~IAlgorithmSelector() noexcept=default
Definition: NvInferRuntimeBase.h:857
Definition: NvInferRuntimeBase.h:469
~IInt8EntropyCalibrator2() noexcept override=default
CalibrationAlgoType getAlgorithm() noexcept override
Definition: NvInfer.h:7705
InterfaceInfo getInterfaceInfo() const noexcept override
Return version information associated with this interface. Applications must not override this method...
Definition: NvInfer.h:7697
Definition: NvInfer.h:7652
CalibrationAlgoType getAlgorithm() noexcept override
Definition: NvInfer.h:7665
InterfaceInfo getInterfaceInfo() const noexcept override
Return version information associated with this interface. Applications must not override this method...
Definition: NvInfer.h:7657
~IInt8EntropyCalibrator() noexcept override=default
Definition: NvInfer.h:7771
CalibrationAlgoType getAlgorithm() noexcept override
Definition: NvInfer.h:7784
InterfaceInfo getInterfaceInfo() const noexcept override
Return version information associated with this interface. Applications must not override this method...
Definition: NvInfer.h:7776
virtual double getQuantile() const noexcept=0
The quantile (between 0 and 1) that will be used to select the region maximum when the quantile metho...
Definition: NvInfer.h:7732
~IInt8MinMaxCalibrator() noexcept override=default
CalibrationAlgoType getAlgorithm() noexcept override
Definition: NvInfer.h:7745
InterfaceInfo getInterfaceInfo() const noexcept override
Return version information associated with this interface. Applications must not override this method...
Definition: NvInfer.h:7737
Definition: NvInferRuntime.h:687
Definition: NvInfer.h:8525
virtual bool stepComplete(char const *phaseName, int32_t step) noexcept=0
Signal that a step of an optimizer phase has finished.
virtual ~IProgressMonitor() noexcept=default
virtual void phaseFinish(char const *phaseName) noexcept=0
Signal that a phase of the optimizer has finished.
virtual void phaseStart(char const *phaseName, char const *parentPhase, int32_t nbSteps) noexcept=0
Signal that a phase of the optimizer has started.
IBuilder * createInferBuilder(ILogger &logger) noexcept
Create an instance of an IBuilder class.
Definition: NvInfer.h:9737
The TensorRT API version 1 namespace.
uint32_t TacticSources
Represents a collection of one or more TacticSource values combine using bitwise-OR operations.
Definition: NvInferRuntime.h:2321
ResizeSelector
The coordinate selector when resize to single pixel output.
Definition: NvInfer.h:3945
@ kFORMULA
Use formula to map the original index.
@ kUPPER
Select the upper left pixel.
EngineCapability
List of supported engine capability flows.
Definition: NvInferRuntime.h:73
nvinfer1::IPluginRegistry * getBuilderPluginRegistry(nvinfer1::EngineCapability capability) noexcept
Return the plugin registry for building a Standard engine, or nullptr if no registry exists.
MemoryPoolType
The type for memory pools used by TensorRT.
Definition: NvInfer.h:8383
ScaleMode
Controls how shift, scale and power are applied in a Scale layer.
Definition: NvInfer.h:1810
@ kUNIFORM
Identical coefficients across all elements of the tensor.
@ kCHANNEL
Per-channel coefficients.
uint32_t QuantizationFlags
Represents one or more QuantizationFlag values using binary OR operations.
Definition: NvInfer.h:8118
HardwareCompatibilityLevel
Describes requirements of compatibility with GPU architectures other than that of the GPU on which th...
Definition: NvInfer.h:8491
BoundingBoxFormat
Representation of bounding box data used for the Boxes input tensor in INMSLayer.
Definition: NvInfer.h:5862
@ kCENTER_SIZES
(x_center, y_center, width, height) where (x_center, y_center) is the center point of the box
@ kCORNER_PAIRS
(x1, y1, x2, y2) where (x1, y1) and (x2, y2) are any pair of diagonal corners
constexpr int32_t EnumMax< BuilderFlag >() noexcept
Definition: NvInfer.h:8300
constexpr int32_t EnumMax< LayerType >() noexcept
Definition: NvInfer.h:114
constexpr int32_t EnumMax< CalibrationAlgoType >() noexcept
Definition: NvInfer.h:7567
UnaryOperation
Enumerates the unary operations that may be performed by a Unary layer.
Definition: NvInfer.h:2758
@ kISINF
Return true if input value equals +/- infinity for floating-point data type.
@ kCOSH
Hyperbolic cosine.
@ kACOSH
Inverse hyperbolic cosine.
@ kERF
Gauss error function.
@ kISNAN
Return true if input value is a NaN for floating-point data type.
@ kACOS
Inverse cosine.
@ kABS
Absolute value.
@ kSINH
Hyperbolic sine.
@ kROUND
Round to nearest even for floating-point data type.
@ kATANH
Inverse hyperbolic tangent.
@ kASINH
Inverse hyperbolic sine.
@ kSIGN
Sign, If input > 0, output 1; if input < 0, output -1; if input == 0, output 0.
@ kEXP
Exponentiation.
@ kATAN
Inverse tangent.
constexpr int32_t EnumMax< ReduceOperation >() noexcept
Definition: NvInfer.h:2867
constexpr int32_t EnumMax< TripLimit >() noexcept
Definition: NvInfer.h:4342
ActivationType
Enumerates the types of activation to perform in an activation layer.
Definition: NvInfer.h:133
@ kSELU
Selu activation: x>0 ? beta * x : beta * (alpha*exp(x) - alpha)
@ kTANH
TanH activation.
@ kSCALED_TANH
Scaled tanh activation: alpha*tanh(beta*x)
@ kRELU
Rectified linear activation.
@ kELU
Elu activation: x>=0 ? x : alpha * (exp(x) - 1).
@ kLEAKY_RELU
LeakyRelu activation: x>=0 ? x : alpha * x.
@ kSOFTSIGN
Softsign activation: x / (1+|x|)
@ kHARD_SIGMOID
Hard sigmoid activation: max(0, min(1, alpha*x+beta))
@ kTHRESHOLDED_RELU
Thresholded ReLU activation: x>alpha ? x : 0.
@ kSIGMOID
Sigmoid activation.
@ kCLIP
Clip activation: max(alpha, min(beta, x))
@ kGELU_TANH
GELU tanh activation: 0.5 * x * (1 + tanh(sqrt(2/pi) * (0.044715F * pow(x, 3) + x)))
@ kGELU_ERF
GELU erf activation: 0.5 * x * (1 + erf(sqrt(0.5) * x))
@ kSOFTPLUS
Parametric softplus activation: alpha*log(exp(beta*x)+1)
FillOperation
Enumerates the tensor fill operations that may performed by a fill layer.
Definition: NvInfer.h:4916
@ kRANDOM_UNIFORM
Randomly draw values from a uniform distribution.
@ kRANDOM_NORMAL
Randomly draw values from a normal distribution.
ResizeRoundMode
The rounding mode for nearest neighbor resize.
Definition: NvInfer.h:3975
@ kHALF_UP
Round half up.
@ kHALF_DOWN
Round half down.
nvinfer1::safe::IPluginRegistry * getBuilderSafePluginRegistry(nvinfer1::EngineCapability capability) noexcept
Return the plugin registry for building a Safety engine, or nullptr if no registry exists.
PaddingMode
Enumerates the modes of padding to perform in convolution, deconvolution and pooling layer,...
Definition: NvInfer.h:988
@ kSAME_LOWER
Use SAME padding, with prePadding >= postPadding.
@ kEXPLICIT_ROUND_DOWN
Use explicit padding, rounding output size down.
@ kEXPLICIT_ROUND_UP
Use explicit padding, rounding output size up.
@ kSAME_UPPER
Use SAME padding, with prePadding <= postPadding.
TripLimit
Definition: NvInfer.h:4330
@ kWHILE
Tensor is a scalar of type kBOOL. Loop terminates when value is false.
@ kCOUNT
Tensor is a scalar of type kINT32 or kINT64 that contains the trip count.
uint32_t NetworkDefinitionCreationFlags
Represents one or more NetworkDefinitionCreationFlag flags using binary OR operations....
Definition: NvInfer.h:9425
PreviewFeature
Define preview features.
Definition: NvInfer.h:8458
constexpr int32_t EnumMax< GatherMode >() noexcept
Definition: NvInfer.h:2518
DataType
The type of weights and tensors.
Definition: NvInferRuntimeBase.h:135
uint32_t BuilderFlags
Represents one or more BuilderFlag values using binary OR operations, e.g., 1U << BuilderFlag::kFP16 ...
Definition: NvInfer.h:8154
DeviceType
The device that this layer/network will execute on.
Definition: NvInferRuntime.h:1273
constexpr int32_t EnumMax< ScaleMode >() noexcept
Definition: NvInfer.h:1822
CalibrationAlgoType
Version of calibration algorithm to use.
Definition: NvInfer.h:7554
@ kENTROPY_CALIBRATION_2
Entropy calibration.
@ kLEGACY_CALIBRATION
Legacy calibration.
@ kENTROPY_CALIBRATION
Legacy entropy calibration.
@ kMINMAX_CALIBRATION
Minmax calibration.
LayerType
The type values of layer classes.
Definition: NvInfer.h:58
@ kGRID_SAMPLE
Grid sample layer.
@ kRAGGED_SOFTMAX
Ragged softmax layer.
@ kDECONVOLUTION
Deconvolution layer.
@ kREDUCE
Reduce layer.
@ kASSERTION
Assertion layer.
@ kTOPK
TopK layer.
@ kRESIZE
Resize Layer.
@ kCAST
Cast layer.
@ kPADDING
Padding layer.
@ kMATRIX_MULTIPLY
Matrix multiply layer.
@ kCONDITION
Condition layer.
@ kCONDITIONAL_INPUT
Conditional Input layer.
@ kIDENTITY
Identity layer.
@ kNORMALIZATION
Normalization layer.
@ kQUANTIZE
Quantize layer.
@ kSCATTER
Scatter layer.
@ kCONVOLUTION
Convolution layer.
@ kPARAMETRIC_RELU
Parametric ReLU layer.
@ kCONCATENATION
Concatenation layer.
@ kONE_HOT
OneHot layer.
@ kREVERSE_SEQUENCE
Reverse sequence layer.
@ kSLICE
Slice layer.
@ kEINSUM
Einsum layer.
@ kSOFTMAX
SoftMax layer.
@ kSHAPE
Shape layer.
@ kRECURRENCE
Loop Recurrence layer.
@ kDEQUANTIZE
Dequantize layer.
@ kSHUFFLE
Shuffle layer.
@ kPLUGIN_V3
PluginV3 layer.
@ kITERATOR
Loop Iterator layer.
@ kPOOLING
Pooling layer.
@ kTRIP_LIMIT
Loop Trip limit layer.
@ kSCALE
Scale layer.
@ kGATHER
Gather layer.
@ kUNARY
UnaryOp operation Layer.
@ kACTIVATION
Activation layer.
@ kELEMENTWISE
Elementwise layer.
@ kSELECT
Select layer.
@ kPLUGIN_V2
PluginV2 layer.
@ kLOOP_OUTPUT
Loop output layer.
@ kCONDITIONAL_OUTPUT
Conditional Output layer.
@ kCONSTANT
Constant layer.
@ kNON_ZERO
NonZero layer.
@ kFILL
Fill layer.
@ kPLUGIN
Plugin layer.
constexpr int32_t EnumMax< QuantizationFlag >() noexcept
Definition: NvInfer.h:8143
SampleMode
Controls how ISliceLayer and IGridSample handle out-of-bounds coordinates.
Definition: NvInfer.h:3214
@ kCLAMP
Out of bounds indices are clamped to bounds.
@ kSTRICT_BOUNDS
Fail with error when the coordinates are out of bounds.
@ kWRAP
Coordinates wrap around periodically.
GatherMode
Control form of IGatherLayer.
Definition: NvInfer.h:2506
@ kDEFAULT
Similar to ONNX Gather.
@ kELEMENT
Similar to ONNX GatherElements.
@ kND
Similar to ONNX GatherND.
uint32_t TensorFormats
It is capable of representing one or more TensorFormat by binary OR operations, e....
Definition: NvInfer.h:125
ProfilingVerbosity
List of verbosity levels of layer information exposed in NVTX annotations and in IEngineInspector.
Definition: NvInferRuntime.h:2333
NetworkDefinitionCreationFlag
List of immutable network properties expressed at network creation time. NetworkDefinitionCreationFla...
Definition: NvInfer.h:9436
ElementWiseOperation
Enumerates the binary operations that may be performed by an ElementWise layer.
Definition: NvInfer.h:2416
@ kSUB
Subtract the second element from the first.
@ kSUM
Sum of the two elements.
@ kPROD
Product of the two elements.
@ kFLOOR_DIV
Floor division of the first element by the second.
@ kEQUAL
Check if two elements are equal.
@ kAND
Logical AND of two elements.
@ kOR
Logical OR of two elements.
@ kMIN
Minimum of the two elements.
@ kPOW
The first element to the power of the second element.
@ kLESS
Check if element in first tensor is less than corresponding element in second tensor.
@ kGREATER
Check if element in first tensor is greater than corresponding element in second tensor.
@ kXOR
Logical XOR of two elements.
@ kDIV
Divide the first element by the second.
QuantizationFlag
List of valid flags for quantizing the network to int8.
Definition: NvInfer.h:8130
constexpr int32_t EnumMax< SampleMode >() noexcept
Definition: NvInfer.h:3230
InterpolationMode
Enumerates various modes of interpolation.
Definition: NvInfer.h:3863
@ kNEAREST
ND (0 < N <= 8) nearest neighbor resizing.
@ kCUBIC
Supports bicubic (2D) interpolation.
@ kLINEAR
Supports linear (1D), bilinear (2D), and trilinear (3D) interpolation.
BuilderFlag
List of valid modes that the builder can enable when creating an engine from a network definition.
Definition: NvInfer.h:8164
@ kWEIGHT_STREAMING
Enable weight streaming for the current engine.
@ kDEBUG
Enable debugging of layers via synchronizing after every layer.
@ kGPU_FALLBACK
Enable layers marked to execute on GPU if layer cannot execute on DLA.
@ kSPARSE_WEIGHTS
Allow the builder to examine weights and use optimized functions when weights have suitable sparsity.
@ kFP16
Enable FP16 layer selection, with FP32 fallback.
@ kINT8
Enable Int8 layer selection, with FP32 fallback with FP16 fallback if kFP16 also specified.
@ kINT4
Enable plugins with INT4 input/output.
@ kSTRIP_PLAN
Strip the refittable weights from the engine plan file.
@ kDISABLE_TIMING_CACHE
Disable reuse of timing information across identical layers.
@ kREFIT
Enable building a refittable engine.
@ kOBEY_PRECISION_CONSTRAINTS
Require that layers execute in specified precisions. Build fails otherwise.
@ kREJECT_EMPTY_ALGORITHMS
Fail if IAlgorithmSelector::selectAlgorithms returns an empty set of algorithms.
constexpr int32_t EnumMax< TopKOperation >() noexcept
Definition: NvInfer.h:3453
constexpr int32_t EnumMax< MemoryPoolType >() noexcept
Definition: NvInfer.h:8444
TopKOperation
Enumerates the operations that may be performed by a TopK layer.
Definition: NvInfer.h:3442
ReduceOperation
Enumerates the reduce operations that may be performed by a Reduce layer.
Definition: NvInfer.h:2853
constexpr int32_t EnumMax< LoopOutput >() noexcept
Definition: NvInfer.h:4321
constexpr int32_t EnumMax< NetworkDefinitionCreationFlag >() noexcept
Definition: NvInfer.h:9455
ScatterMode
Control form of IScatterLayer.
Definition: NvInfer.h:5598
MatrixOperation
Enumerates the operations that may be performed on a tensor by IMatrixMultiplyLayer before multiplica...
Definition: NvInfer.h:3568
@ kTRANSPOSE
Like kNONE, but transpose the matrix dimensions.
ResizeCoordinateTransformation
The resize coordinate transformation function.
Definition: NvInfer.h:3891
constexpr int32_t EnumMax< UnaryOperation >() noexcept
Definition: NvInfer.h:2792
LoopOutput
Definition: NvInfer.h:4304
@ kLAST_VALUE
Output value is value of tensor for last iteration.
@ kCONCATENATE
Output value is concatenation of values of tensor for each iteration, in forward order.
@ kREVERSE
Output value is concatenation of values of tensor for each iteration, in reverse order.
constexpr int32_t EnumMax< BoundingBoxFormat >() noexcept
Definition: NvInfer.h:5875
constexpr int32_t EnumMax< MatrixOperation >() noexcept
Definition: NvInfer.h:3596
PoolingType
The type of pooling to perform in a pooling layer.
Definition: NvInfer.h:1424
@ kAVERAGE
Average over elements. If the tensor is padded, the count includes the padding.
@ kMAX
Maximum over elements.
@ kMAX_AVERAGE_BLEND
Blending between max and average pooling: (1-blendFactor)*maxPool + blendFactor*avgPool.
v_1_0::IProgressMonitor IProgressMonitor
Definition: NvInfer.h:8608
constexpr int32_t EnumMax< FillOperation >() noexcept
Definition: NvInfer.h:4947
TensorLocation
The location for tensor data storage, device or host.
Definition: NvInferRuntime.h:201
OptProfileSelector
When setting or querying optimization profile parameters (such as shape tensor inputs or dynamic dime...
Definition: NvInferRuntime.h:2064
constexpr int32_t EnumMax< ScatterMode >() noexcept
Definition: NvInfer.h:5609
Represents a permutation of dimensions.
Definition: NvInfer.h:3023
Declaration of EnumMaxImpl struct to store maximum number of elements in an enumeration type.
Definition: NvInferRuntimeBase.h:120

  Copyright © 2024 NVIDIA Corporation
  Privacy Policy | Manage My Privacy | Do Not Sell or Share My Data | Terms of Service | Accessibility | Corporate Policies | Product Security | Contact