170 static constexpr int32_t kVALUE = 14;
208 mImpl->setName(name);
220 return mImpl->getName();
239 mImpl->setDimensions(dimensions);
253 return mImpl->getDimensions();
289 mImpl->setType(type);
304 return mImpl->getType();
312 return mImpl->isNetworkInput();
320 return mImpl->isNetworkOutput();
342 mImpl->setAllowedFormats(formats);
355 return mImpl->getAllowedFormats();
386 return mImpl->isShapeTensor();
407 return mImpl->isExecutionTensor();
433 mImpl->setDimensionName(index, name);
448 return mImpl->getDimensionName(index);
473 return mLayer->getType();
487 mLayer->setName(name);
497 return mLayer->getName();
505 return mLayer->getNbInputs();
518 return mLayer->getInput(index);
526 return mLayer->getNbOutputs();
536 return mLayer->getOutput(index);
553 return mLayer->setInput(index, tensor);
586 mLayer->setPrecision(dataType);
598 return mLayer->getPrecision();
612 return mLayer->precisionIsSet();
624 mLayer->resetPrecision();
674 mLayer->setOutputType(index, dataType);
689 return mLayer->getOutputType(index);
705 return mLayer->outputTypeIsSet(index);
719 return mLayer->resetOutputType(index);
737 mLayer->setMetadata(metadata);
750 return mLayer->getMetadata();
771 return mLayer->setNbRanks(nbRanks);
783 return mLayer->getNbRanks();
788 apiv::VLayer* mLayer;
965 static constexpr int32_t kVALUE = 4;
993 mImpl->setNbOutputMaps(nbOutputMaps);
1003 return mImpl->getNbOutputMaps();
1023 mImpl->setNbGroups(nbGroups);
1033 return mImpl->getNbGroups();
1047 mImpl->setKernelWeights(weights);
1057 return mImpl->getKernelWeights();
1072 mImpl->setBiasWeights(weights);
1082 return mImpl->getBiasWeights();
1099 mImpl->setPrePadding(padding);
1109 return mImpl->getPrePadding();
1126 mImpl->setPostPadding(padding);
1136 return mImpl->getPostPadding();
1150 mImpl->setPaddingMode(paddingMode);
1162 return mImpl->getPaddingMode();
1175 mImpl->setKernelSizeNd(kernelSize);
1185 return mImpl->getKernelSizeNd();
1200 mImpl->setStrideNd(stride);
1210 return mImpl->getStrideNd();
1228 mImpl->setPaddingNd(padding);
1240 return mImpl->getPaddingNd();
1254 mImpl->setDilationNd(dilation);
1264 return mImpl->getDilationNd();
1313 mImpl->setActivationType(type);
1323 return mImpl->getActivationType();
1338 mImpl->setAlpha(alpha);
1352 mImpl->setBeta(beta);
1361 return mImpl->getAlpha();
1370 return mImpl->getBeta();
1400 static constexpr int32_t kVALUE = 3;
1427 mImpl->setPoolingType(type);
1437 return mImpl->getPoolingType();
1452 mImpl->setBlendFactor(blendFactor);
1465 return mImpl->getBlendFactor();
1479 mImpl->setAverageCountExcludesPadding(exclusive);
1490 return mImpl->getAverageCountExcludesPadding();
1508 mImpl->setPrePadding(padding);
1518 return mImpl->getPrePadding();
1536 mImpl->setPostPadding(padding);
1546 return mImpl->getPostPadding();
1559 mImpl->setPaddingMode(paddingMode);
1570 return mImpl->getPaddingMode();
1583 mImpl->setWindowSizeNd(windowSize);
1593 return mImpl->getWindowSizeNd();
1608 mImpl->setStrideNd(stride);
1618 return mImpl->getStrideNd();
1637 mImpl->setPaddingNd(padding);
1649 return mImpl->getPaddingNd();
1680 mImpl->setWindowSize(windowSize);
1690 return mImpl->getWindowSize();
1702 mImpl->setAlpha(alpha);
1712 return mImpl->getAlpha();
1724 mImpl->setBeta(beta);
1734 return mImpl->getBeta();
1756 return mImpl->getK();
1822 mImpl->setMode(mode);
1832 return mImpl->getMode();
1842 mImpl->setShift(shift);
1852 return mImpl->getShift();
1862 mImpl->setScale(scale);
1872 return mImpl->getScale();
1882 mImpl->setPower(power);
1892 return mImpl->getPower();
1907 return mImpl->getChannelAxis();
1928 mImpl->setChannelAxis(channelAxis);
1981 mImpl->setAxes(axes);
1991 return mImpl->getAxes();
2027 mImpl->setAxis(axis);
2037 return mImpl->getAxis();
2064 mImpl->setNbOutputMaps(nbOutputMaps);
2074 return mImpl->getNbOutputMaps();
2094 mImpl->setNbGroups(nbGroups);
2104 return mImpl->getNbGroups();
2118 mImpl->setKernelWeights(weights);
2128 return mImpl->getKernelWeights();
2143 mImpl->setBiasWeights(weights);
2153 return mImpl->getBiasWeights();
2170 mImpl->setPrePadding(padding);
2180 return mImpl->getPrePadding();
2197 mImpl->setPostPadding(padding);
2207 return mImpl->getPostPadding();
2221 mImpl->setPaddingMode(paddingMode);
2233 return mImpl->getPaddingMode();
2248 mImpl->setKernelSizeNd(kernelSize);
2258 return mImpl->getKernelSizeNd();
2275 mImpl->setStrideNd(stride);
2285 return mImpl->getStrideNd();
2303 mImpl->setPaddingNd(padding);
2315 return mImpl->getPaddingNd();
2341 mImpl->setDilationNd(dilation);
2351 return mImpl->getDilationNd();
2399 static constexpr int32_t kVALUE = 14;
2436 return mImpl->setOperation(op);
2448 return mImpl->getOperation();
2569 mImpl->setGatherAxis(axis);
2581 return mImpl->getGatherAxis();
2604 mImpl->setNbElementWiseDims(elementWiseDims);
2614 return mImpl->getNbElementWiseDims();
2624 mImpl->setMode(mode);
2634 return mImpl->getMode();
2663 return mImpl->getPlugin();
2690 return mImpl->getPlugin();
2773 mImpl->setOperation(op);
2783 return mImpl->getOperation();
2859 static constexpr int32_t kVALUE = 5;
2879 mImpl->setOperation(op);
2889 return mImpl->getOperation();
2899 mImpl->setReduceAxes(reduceAxes);
2909 return mImpl->getReduceAxes();
2919 mImpl->setKeepDimensions(keepDimensions);
2929 return mImpl->getKeepDimensions();
2963 mImpl->setPrePaddingNd(padding);
2975 return mImpl->getPrePaddingNd();
2989 mImpl->setPostPaddingNd(padding);
3001 return mImpl->getPostPaddingNd();
3051 mImpl->setFirstTranspose(permutation);
3063 return mImpl->getFirstTranspose();
3091 mImpl->setReshapeDimensions(dimensions);
3104 return mImpl->getReshapeDimensions();
3151 mImpl->setSecondTranspose(permutation);
3163 return mImpl->getSecondTranspose();
3179 return mImpl->setZeroIsPlaceholder(zeroIsPlaceholder);
3192 return mImpl->getZeroIsPlaceholder();
3303 mImpl->setStart(start);
3318 return mImpl->getStart();
3332 return mImpl->setSize(size);
3347 return mImpl->getSize();
3361 mImpl->setStride(stride);
3376 return mImpl->getStride();
3386 mImpl->setMode(mode);
3396 return mImpl->getMode();
3439 mImpl->setAxes(axes);
3454 return mImpl->getAxes();
3524 mImpl->setOperation(op);
3534 return mImpl->getOperation();
3562 return mImpl->getK();
3572 mImpl->setReduceAxes(reduceAxes);
3582 return mImpl->getReduceAxes();
3613 return mImpl->setIndicesType(type);
3625 return mImpl->getIndicesType();
3711 mImpl->setOperation(index, op);
3723 return mImpl->getOperation(index);
3767 return mImpl->setIndicesType(type);
3779 return mImpl->getIndicesType();
3876 mImpl->setToType(toType);
3887 return mImpl->getToType();
3916 mImpl->setWeights(weights);
3926 return mImpl->getWeights();
3938 mImpl->setDimensions(dimensions);
3950 return mImpl->getDimensions();
3996 static constexpr int32_t kVALUE = 3;
4050 static constexpr int32_t kVALUE = 3;
4080 static constexpr int32_t kVALUE = 2;
4116 static constexpr int32_t kVALUE = 4;
4179 return mImpl->setOutputDimensions(dimensions);
4189 return mImpl->getOutputDimensions();
4217 void setScales(
float const* scales, int32_t nbScales)
noexcept
4219 mImpl->setScales(scales, nbScales);
4236 int32_t
getScales(int32_t size,
float* scales)
const noexcept
4238 return mImpl->getScales(size, scales);
4250 mImpl->setResizeMode(interpolationMode);
4260 return mImpl->getResizeMode();
4295 mImpl->setCoordinateTransformation(coordTransform);
4305 return mImpl->getCoordinateTransformation();
4320 mImpl->setSelectorForSinglePixel(selector);
4330 return mImpl->getSelectorForSinglePixel();
4344 mImpl->setNearestRounding(value);
4354 return mImpl->getNearestRounding();
4376 mImpl->setCubicCoeff(A);
4386 return mImpl->getCubicCoeff();
4399 mImpl->setExcludeOutside(excludeFlag);
4409 return mImpl->getExcludeOutside();
4491 return mBoundary->getLoop();
4496 apiv::VLoopBoundaryLayer* mBoundary;
4514 return mBoundary->getConditional();
4519 apiv::VConditionalBoundaryLayer* mBoundary;
4603 return mImpl->setCondition(condition);
4621 return mImpl->addOutput(trueSubgraphOutput, falseSubgraphOutput);
4633 return mImpl->addInput(input);
4648 mImpl->setName(name);
4658 return mImpl->getName();
4728 return mImpl->getLoopOutput();
4745 mImpl->setAxis(axis);
4753 return mImpl->getAxis();
4802 return mImpl->getTripLimit();
4828 mImpl->setAxis(axis);
4836 return mImpl->getAxis();
4850 mImpl->setReverse(reverse);
4860 return mImpl->getReverse();
4889 return mImpl->addRecurrence(initialValue);
4910 return mImpl->addTripLimit(tensor, limit);
4923 return mImpl->addIterator(tensor, axis, reverse);
4936 return mImpl->addLoopOutput(tensor, outputKind, axis);
4951 mImpl->setName(name);
4961 return mImpl->getName();
5016 mImpl->setMessage(message);
5026 return mImpl->getMessage();
5131 mImpl->setDimensions(dimensions);
5146 return mImpl->getDimensions();
5156 mImpl->setOperation(op);
5166 return mImpl->getOperation();
5185 mImpl->setAlpha(alpha);
5200 return mImpl->getAlpha();
5219 mImpl->setBeta(beta);
5234 return mImpl->getBeta();
5295 mImpl->setAlphaInt64(alpha);
5310 return mImpl->getAlphaInt64();
5329 mImpl->setBetaInt64(beta);
5344 return mImpl->getBetaInt64();
5352 return mImpl->isAlphaBetaInt64();
5370 mImpl->setToType(toType);
5382 return mImpl->getToType();
5477 return mImpl->getAxis();
5488 mImpl->setAxis(axis);
5501 return mImpl->setBlockShape(blockShape);
5512 return mImpl->getBlockShape();
5528 mImpl->setToType(toType);
5540 return mImpl->getToType();
5629 return mImpl->getAxis();
5640 mImpl->setAxis(axis);
5657 return mImpl->setBlockShape(blockShape);
5668 return mImpl->getBlockShape();
5684 mImpl->setToType(toType);
5696 return mImpl->getToType();
5751 mImpl->setToType(toType);
5764 return mImpl->getToType();
5777 mImpl->setScaleType(scaleType);
5790 return mImpl->getScaleType();
5803 mImpl->setAxis(axis);
5813 return mImpl->getAxis();
5826 mImpl->setBlockSize(size);
5836 return mImpl->getBlockSize();
5849 mImpl->setBlockShape(blockShape);
5861 return mImpl->getBlockShape();
5917 return mImpl->setEquation(equation);
5927 return mImpl->getEquation();
6026 mImpl->setMode(mode);
6036 return mImpl->getMode();
6046 mImpl->setAxis(axis);
6054 return mImpl->getAxis();
6098 mImpl->setAxis(axis);
6106 return mImpl->getAxis();
6135 mImpl->setInterpolationMode(mode);
6147 return mImpl->getInterpolationMode();
6157 mImpl->setAlignCorners(alignCorners);
6169 return mImpl->getAlignCorners();
6181 return mImpl->setSampleMode(mode);
6193 return mImpl->getSampleMode();
6291 mImpl->setBoundingBoxFormat(fmt);
6303 return mImpl->getBoundingBoxFormat();
6317 mImpl->setTopKBoxLimit(limit);
6327 return mImpl->getTopKBoxLimit();
6362 return mImpl->setIndicesType(type);
6374 return mImpl->getIndicesType();
6407 mImpl->setBatchAxis(batchAxis);
6417 return mImpl->getBatchAxis();
6430 mImpl->setSequenceAxis(sequenceAxis);
6440 return mImpl->getSequenceAxis();
6478 return mImpl->setEpsilon(eps);
6488 return mImpl->getEpsilon();
6498 return mImpl->setAxes(axesMask);
6508 return mImpl->getAxes();
6529 return mImpl->setNbGroups(nbGroups);
6539 return mImpl->getNbGroups();
6567 return mImpl->setComputePrecision(type);
6579 return mImpl->getComputePrecision();
6589 return mImpl->isV2();
6686 static constexpr int32_t kVALUE = 1;
6732 return mImpl->setOperation(op);
6744 return mImpl->getOperation();
6756 mImpl->setExclusive(exclusive);
6768 return mImpl->getExclusive();
6780 mImpl->setReverse(reverse);
6792 return mImpl->getReverse();
6822 static constexpr int32_t kVALUE = 2;
6844 return mBoundary->getAttention();
6849 apiv::VAttentionBoundaryLayer* mBoundary;
6966 return mImpl->setNormalizationOperation(op);
6978 return mImpl->getNormalizationOperation();
6995 return mImpl->setMask(mask);
7007 return mImpl->getMask();
7020 return mImpl->setCausal(isCausal);
7032 return mImpl->getCausal();
7044 return mImpl->setDecomposable(decomposable);
7057 return mImpl->getDecomposable();
7076 return mImpl->setInput(index, input);
7085 return mImpl->getNbInputs();
7097 return mImpl->getInput(index);
7105 return mImpl->getNbOutputs();
7117 return mImpl->getOutput(index);
7134 return mImpl->setName(name);
7146 return mImpl->getName();
7162 return mImpl->setNormalizationQuantizeScale(tensor);
7173 return mImpl->getNormalizationQuantizeScale();
7186 return mImpl->setNormalizationQuantizeToType(type);
7198 return mImpl->getNormalizationQuantizeToType();
7218 return mImpl->setMetadata(metadata);
7231 return mImpl->getMetadata();
7247 return mImpl->setNbRanks(nbRanks);
7259 return mImpl->getNbRanks();
7283 mImpl->setInterleaved(interleaved);
7294 return mImpl->getInterleaved();
7305 return mImpl->setRotaryEmbeddingDim(rotaryEmbeddingDim);
7316 return mImpl->getRotaryEmbeddingDim();
7361 static constexpr int32_t kVALUE = 1;
7411 return mImpl->setCacheMode(cacheMode);
7421 return mImpl->getCacheMode();
7451 static constexpr int32_t kVALUE = 2;
7583 mImpl->setGatedWeights(fcGateWeights, fcUpWeights, fcDownWeights, activationType);
7595 mImpl->setGatedBiases(fcGateBiases, fcUpBiases, fcDownBiases);
7607 mImpl->setActivationType(activationType);
7619 return mImpl->getActivationType();
7642 mImpl->setQuantizationStatic(fcDownActivationScale, dataType);
7671 mImpl->setQuantizationDynamicDblQ(fcDownActivationDblQScale, dataType, blockShape, dynQOutputScaleType);
7686 mImpl->setQuantizationToType(type);
7698 return mImpl->getQuantizationToType();
7714 mImpl->setQuantizationBlockShape(blockShape);
7726 return mImpl->getQuantizationBlockShape();
7738 mImpl->setDynQOutputScaleType(type);
7750 return mImpl->getDynQOutputScaleType();
7771 mImpl->setSwigluParams(limit, alpha, beta);
7785 mImpl->setSwigluParamLimit(limit);
7797 return mImpl->getSwigluParamLimit();
7811 mImpl->setSwigluParamAlpha(alpha);
7823 return mImpl->getSwigluParamAlpha();
7837 mImpl->setSwigluParamBeta(beta);
7849 return mImpl->getSwigluParamBeta();
7866 mImpl->setInput(index, tensor);
7949 return mImpl->addInput(name, type, dimensions);
7963 mImpl->markOutput(tensor);
7981 return mImpl->markDebug(tensor);
7997 return mImpl->unmarkDebug(tensor);
8007 return mImpl->isDebugTensor(tensor);
8029 return mImpl->markUnfusedTensorsAsDebugTensors();
8043 return mImpl->unmarkUnfusedTensorsAsDebugTensors();
8063 return mImpl->addActivation(input, type);
8082 return mImpl->addLRN(input, window, alpha, beta, k);
8108 return mImpl->addScale(input, mode, shift, scale, power);
8121 return mImpl->addSoftMax(input);
8138 return mImpl->addConcatenation(inputs, nbInputs);
8165 return mImpl->addElementWise(input1, input2, op);
8187 return mImpl->addUnary(input, operation);
8201 return mImpl->addShuffle(input);
8218 return mImpl->addOneHot(indices, values, depth, axis);
8230 return mImpl->getNbLayers();
8244 return mImpl->getLayer(index);
8256 return mImpl->getNbInputs();
8272 return mImpl->getInput(index);
8286 return mImpl->getNbOutputs();
8302 return mImpl->getOutput(index);
8329 return mImpl->addReduce(input, operation, reduceAxes, keepDimensions);
8364 return mImpl->addTopK(input, op, k, reduceAxes);
8397 return mImpl->addTopKV2(input, op, k, reduceAxes, indicesType);
8413 return mImpl->addGather(data, indices, axis);
8429 return mImpl->addGatherV2(data, indices, mode);
8448 return mImpl->addRaggedSoftMax(input, bounds);
8470 return mImpl->addMatrixMultiply(input0, op0, input1, op1);
8488 return mImpl->addNonZero(input);
8504 return mImpl->addNonZeroV2(input, indicesType);
8528 return mImpl->addConstant(dimensions, weights);
8542 return mImpl->addIdentity(input);
8557 return mImpl->addCast(input, toType);
8572 mImpl->removeTensor(tensor);
8584 mImpl->unmarkOutput(tensor);
8603 return mImpl->addSlice(input, start, size, stride);
8627 mImpl->setName(name);
8641 return mImpl->getName();
8657 return mImpl->addShape(input);
8667 return mImpl->getFlags();
8679 return mImpl->getFlag(networkDefinitionCreationFlag);
8696 return mImpl->markOutputForShapes(tensor);
8708 return mImpl->unmarkOutputForShapes(tensor);
8726 return mImpl->addParametricReLU(input, slope);
8749 return mImpl->addConvolutionNd(input, nbOutputMaps, kernelSize, kernelWeights, biasWeights);
8768 return mImpl->addPoolingNd(input, type, windowSize);
8791 return mImpl->addDeconvolutionNd(input, nbOutputMaps, kernelSize, kernelWeights, biasWeights);
8828 return mImpl->addScaleNd(input, mode, shift, scale, power, channelAxis);
8844 return mImpl->addResize(input);
8858 return mImpl->addLoop();
8873 return mImpl->addIfConditional();
8912 return mImpl->addSelect(condition, thenInput, elseInput);
8929 return mImpl->addAssertion(condition, message);
8955 return mImpl->addFillV2(dimensions, op, outputType);
8971 return mImpl->addPaddingNd(input, prePadding, postPadding);
8995 return mImpl->setWeightsName(weights, name);
9014 mImpl->setErrorRecorder(recorder);
9029 return mImpl->getErrorRecorder();
9052 return mImpl->addDequantizeV2(input, scale, outputType);
9072 return mImpl->addScatter(data, indices, updates, mode);
9096 return mImpl->addQuantizeV2(input, scale, outputType);
9124 return mImpl->addDynamicQuantize(input, axis, blockSize, outputType, scaleType);
9148 return mImpl->addDynamicQuantizeV2(input, blockShape, outputType, scaleType);
9163 return mImpl->addEinsum(inputs, nbInputs, equation);
9181 return mImpl->addGridSample(input, grid);
9203 return mImpl->addNMS(boxes, scores, maxOutputBoxesPerClass);
9223 return mImpl->addNMSV2(boxes, scores, maxOutputBoxesPerClass, indicesType);
9240 return mImpl->addReverseSequence(input, sequenceLens);
9272 return mImpl->addNormalization(input, scale, bias, axesMask);
9294 return mImpl->addCumulative(input, axis, operation, exclusive, reverse);
9322 return mImpl->addAttention(query, key, value, normOp, causal);
9346 return mImpl->addRotaryEmbedding(input, cosCache, sinCache, interleaved, rotaryEmbeddingDim);
9381 return mImpl->addKVCacheUpdate(cache, update, writeIndices, cacheMode);
9400 return mImpl->addMoE(hiddenStates, selectedExpertsForTokens, scoresForSelectedExperts);
9427 ReduceOperation reduceOp, int64_t root, int64_t* groups, int64_t groupSize)
noexcept
9429 return mImpl->addDistCollective(input, distCollectiveOp, reduceOp, root, groups, groupSize);
9440 return mImpl->getBuilder();
9453 return mImpl->markWeightsRefittable(name);
9465 return mImpl->unmarkWeightsRefittable(name);
9478 return mImpl->areWeightsMarkedRefittable(name);
9497 return mImpl->addSqueeze(input, axes);
9518 return mImpl->addUnsqueeze(input, axes);
9544 return mImpl->addNormalizationV2(input, scale, bias, axesMask);
9591 static constexpr int32_t kVALUE = 2;
9790#if ENABLE_FEATURE_DISABLE_RUNTIME_ALLOCATION
9797 kREQUIRE_USER_ALLOCATION = 29,
9810#if ENABLE_FEATURE_DISABLE_RUNTIME_ALLOCATION
9854 static constexpr uint64_t kINVALID_TACTIC_HASH = UINT64_MAX;
9889 return mImpl->serialize();
9913 return mImpl->combine(inputCache, ignoreMismatch);
9923 return mImpl->reset();
9940 int64_t
queryKeys(TimingCacheKey* keyBuffer, int64_t capacity)
const noexcept
9942 return mImpl->queryKeys(keyBuffer, capacity);
9957 TimingCacheValue
query(TimingCacheKey
const& key)
const noexcept
9959 return mImpl->query(key);
9979 bool update(TimingCacheKey
const& key, TimingCacheValue
const& value)
noexcept
9981 return mImpl->update(key, value);
10111 static constexpr int32_t kVALUE = 4;
10163 static constexpr int32_t kVALUE = 3;
10225 static constexpr int32_t kVALUE = 4;
10264 virtual void phaseStart(
char const* phaseName,
char const* parentPhase, int32_t nbSteps)
noexcept = 0;
10337 virtual
void setAvgTimingIterations(int32_t avgTiming) noexcept
10339 mImpl->setAvgTimingIterations(avgTiming);
10351 return mImpl->getAvgTimingIterations();
10364 mImpl->setEngineCapability(capability);
10376 return mImpl->getEngineCapability();
10393 mImpl->setFlags(builderFlags);
10405 return mImpl->getFlags();
10417 mImpl->clearFlag(builderFlag);
10429 mImpl->setFlag(builderFlag);
10441 return mImpl->getFlag(builderFlag);
10458 mImpl->setDeviceType(layer, deviceType);
10468 return mImpl->getDeviceType(layer);
10480 return mImpl->isDeviceTypeSet(layer);
10490 mImpl->resetDeviceType(layer);
10500 return mImpl->canRunOnDLA(layer);
10516 mImpl->setDLACore(dlaCore);
10526 return mImpl->getDLACore();
10537 mImpl->setDefaultDeviceType(deviceType);
10547 return mImpl->getDefaultDeviceType();
10569 return mImpl->setProfileStream(stream);
10581 return mImpl->getProfileStream();
10598 return mImpl->addOptimizationProfile(profile);
10611 return mImpl->getNbOptimizationProfiles();
10623 mImpl->setProfilingVerbosity(verbosity);
10636 return mImpl->getProfilingVerbosity();
10658 return mImpl->setTacticSources(tacticSources);
10673 return mImpl->getTacticSources();
10695 return mImpl->createTimingCache(blob, size);
10720 return mImpl->setTimingCache(cache, ignoreMismatch);
10732 return mImpl->getTimingCache();
10764 mImpl->setMemoryPoolLimit(pool, poolSize);
10783 return mImpl->getMemoryPoolLimit(pool);
10801 mImpl->setPreviewFeature(feature, enable);
10815 return mImpl->getPreviewFeature(feature);
10848 mImpl->setBuilderOptimizationLevel(level);
10860 return mImpl->getBuilderOptimizationLevel();
10877 mImpl->setHardwareCompatibilityLevel(hardwareCompatibilityLevel);
10890 return mImpl->getHardwareCompatibilityLevel();
10903 mImpl->setPluginsToSerialize(paths, nbPaths);
10916 return mImpl->getPluginToSerialize(index);
10926 return mImpl->getNbPluginsToSerialize();
10955 mImpl->setMaxAuxStreams(nbStreams);
10965 return mImpl->getMaxAuxStreams();
10981 return mImpl->setProgressMonitor(monitor);
10991 return mImpl->getProgressMonitor();
11007 mImpl->setRuntimePlatform(runtimePlatform);
11019 return mImpl->getRuntimePlatform();
11031 mImpl->setMaxNbTactics(maxNbTactics);
11043 return mImpl->getMaxNbTactics();
11059 return mImpl->setTilingOptimizationLevel(level);
11071 return mImpl->getTilingOptimizationLevel();
11087 return mImpl->setL2LimitForTiling(size);
11099 return mImpl->getL2LimitForTiling();
11118 return mImpl->setNbComputeCapabilities(maxNbComputeCapabilities);
11130 return mImpl->getNbComputeCapabilities();
11148 return mImpl->setComputeCapability(computeCapability, index);
11162 return mImpl->getComputeCapability(index);
11230 int32_t getMaxDLABatchSize() const noexcept
11232 return mImpl->getMaxDLABatchSize();
11240 return mImpl->getNbDLACores();
11258 mImpl->setGpuAllocator(allocator);
11272 return mImpl->createBuilderConfig();
11298 return mImpl->createNetworkV2(flags);
11313 return mImpl->createOptimizationProfile();
11332 mImpl->setErrorRecorder(recorder);
11347 return mImpl->getErrorRecorder();
11374 return mImpl->buildSerializedNetwork(network, config);
11396 return mImpl->buildSerializedNetworkToStream(network, config, writer);
11423 return mImpl->isNetworkSupported(network, config);
11433 return mImpl->getLogger();
11449 return mImpl->setMaxThreads(maxThreads);
11463 return mImpl->getMaxThreads();
11473 return mImpl->getPluginRegistry();
11486extern "C" TENSORRTAPI void* createInferBuilder_INTERNAL(
void* logger, int32_t version)
noexcept;
#define TENSORRTAPI
Definition: NvInferRuntimeBase.h:70
#define NV_TENSORRT_VERSION
Definition: NvInferRuntimeBase.h:102
#define TRT_NODISCARD
A stand-in for [[nodiscard]] and [[nodiscard(REASON)]] that works with older compilers.
Definition: NvInferRuntimeBase.h:57
#define TRT_DEPRECATED
Definition: NvInferRuntimeBase.h:42
#define TRT_DEPRECATED_ENUM
Definition: NvInferRuntimeBase.h:43
Definition: NvInferRuntimeBase.h:219
static constexpr int32_t MAX_DIMS
The maximum rank (number of dimensions) supported for a tensor.
Definition: NvInferRuntimeBase.h:222
An Activation layer in a network definition.
Definition: NvInfer.h:1302
void setBeta(float beta) noexcept
Set the beta parameter (must be finite).
Definition: NvInfer.h:1350
void setActivationType(ActivationType type) noexcept
Set the type of activation to be performed.
Definition: NvInfer.h:1311
ActivationType getActivationType() const noexcept
Get the type of activation to be performed.
Definition: NvInfer.h:1321
float getAlpha() const noexcept
Get the alpha parameter.
Definition: NvInfer.h:1359
virtual ~IActivationLayer() noexcept=default
float getBeta() const noexcept
Get the beta parameter.
Definition: NvInfer.h:1368
void setAlpha(float alpha) noexcept
Set the alpha parameter (must be finite).
Definition: NvInfer.h:1336
An assertion layer in a network.
Definition: NvInfer.h:5004
void setMessage(char const *message) noexcept
Set the message to print if the assertion fails.
Definition: NvInfer.h:5014
char const * getMessage() const noexcept
Return the assertion message.
Definition: NvInfer.h:5024
virtual ~IAssertionLayer() noexcept=default
This is a base class for Attention boundary layers.
Definition: NvInfer.h:6837
IAttention * getAttention() const noexcept
Get a pointer to the IAttention associated with this boundary layer.
Definition: NvInfer.h:6842
virtual ~IAttentionBoundaryLayer() noexcept=default
Helper for constructing an attention that consumes query, key and value tensors.
Definition: NvInfer.h:6955
ITensor * getMask() noexcept
Get the optional mask in attention.
Definition: NvInfer.h:7005
bool setMetadata(char const *metadata) noexcept
Set the metadata for IAttention.
Definition: NvInfer.h:7216
bool setDecomposable(bool decomposable) noexcept
Set whether the attention can be decomposed to use multiple kernels if no fused kernel support found.
Definition: NvInfer.h:7042
bool setName(char const *name) noexcept
Set the name of the attention.
Definition: NvInfer.h:7132
bool getDecomposable() const noexcept
Get whether the attention can be decomposed to use multiple kernels if no fused kernel support found.
Definition: NvInfer.h:7055
ITensor * getInput(int32_t index) const noexcept
Get the IAttention input corresponding to the given index.
Definition: NvInfer.h:7095
ITensor * getOutput(int32_t index) const noexcept
Get the IAttention output corresponding to the given index. IAttention has only one output.
Definition: NvInfer.h:7115
int32_t getNbOutputs() const noexcept
Get the number of outputs of a layer. IAttention has one output.
Definition: NvInfer.h:7103
bool setNbRanks(int32_t nbRanks) noexcept
Set the number of ranks for multi-device attention execution.
Definition: NvInfer.h:7245
int32_t getNbInputs() const noexcept
Get the number of inputs of IAttention. IAttention has three inputs.
Definition: NvInfer.h:7083
bool setCausal(bool isCausal) noexcept
Set whether the attention will run a causal inference. Cannot be used together with setMask().
Definition: NvInfer.h:7018
bool setNormalizationOperation(AttentionNormalizationOp op) noexcept
Set the normalization operation for the attention.
Definition: NvInfer.h:6964
char const * getName() const noexcept
Return the name of the attention.
Definition: NvInfer.h:7144
bool setNormalizationQuantizeToType(DataType type) noexcept
Set the datatype the attention normalization is quantized to.
Definition: NvInfer.h:7184
int32_t getNbRanks() const noexcept
Get the number of ranks for multi-device execution.
Definition: NvInfer.h:7257
AttentionNormalizationOp getNormalizationOperation() const noexcept
Get the normalization operation for the attention.
Definition: NvInfer.h:6976
bool setNormalizationQuantizeScale(ITensor &tensor) noexcept
Set the quantization scale for the attention normalization output.
Definition: NvInfer.h:7160
char const * getMetadata() const noexcept
Get the metadata of IAttention.
Definition: NvInfer.h:7229
DataType getNormalizationQuantizeToType() const noexcept
Get the datatype the attention normalization is quantized to.
Definition: NvInfer.h:7196
ITensor * getNormalizationQuantizeScale() const noexcept
Get the quantization scale for the attention normalization output.
Definition: NvInfer.h:7171
bool setInput(int32_t index, ITensor &input) noexcept
Append or replace an input of this layer with a specific tensor.
Definition: NvInfer.h:7074
bool setMask(ITensor &mask) noexcept
Set whether a mask will be used for the normalization operation.
Definition: NvInfer.h:6993
bool getCausal() const noexcept
Get whether the attention will run a causal inference.
Definition: NvInfer.h:7030
apiv::VAttention * mImpl
Definition: NvInfer.h:7263
virtual ~IAttention() noexcept=default
This layer represents an output of an IAttention.
Definition: NvInfer.h:6898
virtual ~IAttentionOutputLayer() noexcept=default
Holds properties for configuring a builder to produce an engine.
Definition: NvInfer.h:10325
void setMemoryPoolLimit(MemoryPoolType pool, std::size_t poolSize) noexcept
Set the memory size for the memory pool.
Definition: NvInfer.h:10762
bool setComputeCapability(ComputeCapability computeCapability, int32_t index) noexcept
Set one compute capability for runtime execution.
Definition: NvInfer.h:11146
bool setNbComputeCapabilities(int32_t maxNbComputeCapabilities) noexcept
Set the number of compute capabilities.
Definition: NvInfer.h:11116
TRT_DEPRECATED bool setTimingCache(ITimingCache const &cache, bool ignoreMismatch) noexcept
Attach a timing cache to IBuilderConfig.
Definition: NvInfer.h:10718
void setPreviewFeature(PreviewFeature feature, bool enable) noexcept
Enable or disable a specific preview feature.
Definition: NvInfer.h:10799
bool getPreviewFeature(PreviewFeature feature) const noexcept
Get status of preview feature.
Definition: NvInfer.h:10813
int32_t getBuilderOptimizationLevel() noexcept
Get builder optimization level.
Definition: NvInfer.h:10858
bool setTacticSources(TacticSources tacticSources) noexcept
Set tactic sources.
Definition: NvInfer.h:10656
void setPluginsToSerialize(char const *const *paths, int32_t nbPaths) noexcept
Set the plugin libraries to be serialized with version-compatible engines.
Definition: NvInfer.h:10901
bool setTilingOptimizationLevel(TilingOptimizationLevel level) noexcept
Set the Tiling optimization level.
Definition: NvInfer.h:11057
bool setL2LimitForTiling(int64_t size) noexcept
Set the L2 cache usage limit for Tiling optimization.
Definition: NvInfer.h:11085
std::size_t getMemoryPoolLimit(MemoryPoolType pool) const noexcept
Get the memory size limit of the memory pool.
Definition: NvInfer.h:10781
int32_t getDLACore() const noexcept
Get the DLA core that the engine executes on.
Definition: NvInfer.h:10524
int32_t getNbPluginsToSerialize() const noexcept
Get the number of plugin library paths to be serialized with version-compatible engines.
Definition: NvInfer.h:10924
void setDeviceType(ILayer const *layer, DeviceType deviceType) noexcept
Set the device that this layer must execute on.
Definition: NvInfer.h:10456
void setEngineCapability(EngineCapability capability) noexcept
Configure the builder to target specified EngineCapability flow.
Definition: NvInfer.h:10362
int32_t getMaxAuxStreams() const noexcept
Get the maximum number of auxiliary streams that TRT is allowed to use.
Definition: NvInfer.h:10963
bool getFlag(BuilderFlag builderFlag) const noexcept
Returns true if the build mode flag is set.
Definition: NvInfer.h:10439
void setMaxNbTactics(int32_t maxNbTactics) noexcept
Set the maximum number of tactics to time when there is a choice of tactics.
Definition: NvInfer.h:11029
int64_t getL2LimitForTiling() const noexcept
Get the L2 cache usage limit for tiling optimization.
Definition: NvInfer.h:11097
void setProgressMonitor(IProgressMonitor *monitor) noexcept
Sets the progress monitor for building a network.
Definition: NvInfer.h:10979
void setProfilingVerbosity(ProfilingVerbosity verbosity) noexcept
Set verbosity level of layer information exposed in NVTX annotations and IEngineInspector.
Definition: NvInfer.h:10621
int32_t getNbOptimizationProfiles() const noexcept
Get number of optimization profiles.
Definition: NvInfer.h:10609
void reset() noexcept
Resets the builder configuration to defaults.
Definition: NvInfer.h:10555
char const * getPluginToSerialize(int32_t index) const noexcept
Get the plugin library path to be serialized with version-compatible engines.
Definition: NvInfer.h:10914
EngineCapability getEngineCapability() const noexcept
Query EngineCapability flow configured for the builder.
Definition: NvInfer.h:10374
RuntimePlatform getRuntimePlatform() const noexcept
Get the target platform for runtime execution.
Definition: NvInfer.h:11017
DeviceType getDefaultDeviceType() const noexcept
Get the default DeviceType which was set by setDefaultDeviceType.
Definition: NvInfer.h:10545
void setRuntimePlatform(RuntimePlatform runtimePlatform) noexcept
Set the target platform for runtime execution.
Definition: NvInfer.h:11005
int32_t getMaxNbTactics() const noexcept
Query the maximum number of tactics timed when there is a choice.
Definition: NvInfer.h:11041
BuilderFlags getFlags() const noexcept
Get the build mode flags for this builder config. Defaults to 0.
Definition: NvInfer.h:10403
void setFlags(BuilderFlags builderFlags) noexcept
Set the build mode flags to turn on builder options for this network.
Definition: NvInfer.h:10391
TacticSources getTacticSources() const noexcept
Get tactic sources.
Definition: NvInfer.h:10671
void resetDeviceType(ILayer const *layer) noexcept
reset the DeviceType for this layer
Definition: NvInfer.h:10488
ComputeCapability getComputeCapability(int32_t index) const noexcept
Get one compute capability for runtime execution.
Definition: NvInfer.h:11160
void setDLACore(int32_t dlaCore) noexcept
Sets the DLA core used by the network. Defaults to -1.
Definition: NvInfer.h:10514
HardwareCompatibilityLevel getHardwareCompatibilityLevel() const noexcept
Get the hardware compatibility level.
Definition: NvInfer.h:10888
int32_t getNbComputeCapabilities() const noexcept
Get the number of compute capabilities.
Definition: NvInfer.h:11128
void clearFlag(BuilderFlag builderFlag) noexcept
clear a single build mode flag.
Definition: NvInfer.h:10415
int32_t addOptimizationProfile(IOptimizationProfile const *profile) noexcept
Add an optimization profile.
Definition: NvInfer.h:10596
IProgressMonitor * getProgressMonitor() const noexcept
Definition: NvInfer.h:10989
apiv::VBuilderConfig * mImpl
Definition: NvInfer.h:11166
int32_t getAvgTimingIterations() const noexcept
Query the number of averaging iterations.
Definition: NvInfer.h:10349
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:10535
void setFlag(BuilderFlag builderFlag) noexcept
Set a single build mode flag.
Definition: NvInfer.h:10427
TRT_DEPRECATED nvinfer1::ITimingCache * createTimingCache(void const *blob, std::size_t size) const noexcept
Create timing cache.
Definition: NvInfer.h:10693
virtual ~IBuilderConfig() noexcept=default
DeviceType getDeviceType(ILayer const *layer) const noexcept
Get the device that this layer executes on.
Definition: NvInfer.h:10466
bool canRunOnDLA(ILayer const *layer) const noexcept
Checks if a layer can run on DLA.
Definition: NvInfer.h:10498
cudaStream_t getProfileStream() const noexcept
Get the CUDA stream that is used to profile this network.
Definition: NvInfer.h:10579
void setHardwareCompatibilityLevel(HardwareCompatibilityLevel hardwareCompatibilityLevel) noexcept
Set the hardware compatibility level.
Definition: NvInfer.h:10875
TilingOptimizationLevel getTilingOptimizationLevel() const noexcept
Get the Tiling optimization level.
Definition: NvInfer.h:11069
void setMaxAuxStreams(int32_t nbStreams) noexcept
Set the maximum number of auxiliary streams that TRT is allowed to use.
Definition: NvInfer.h:10953
ProfilingVerbosity getProfilingVerbosity() const noexcept
Get verbosity level of layer information exposed in NVTX annotations and IEngineInspector.
Definition: NvInfer.h:10634
TRT_DEPRECATED nvinfer1::ITimingCache const * getTimingCache() const noexcept
Get the pointer to the timing cache from current IBuilderConfig.
Definition: NvInfer.h:10730
bool isDeviceTypeSet(ILayer const *layer) const noexcept
whether the DeviceType has been explicitly set for this layer
Definition: NvInfer.h:10478
void setBuilderOptimizationLevel(int32_t level) noexcept
Set builder optimization level.
Definition: NvInfer.h:10846
void setProfileStream(const cudaStream_t stream) noexcept
Set the CUDA stream that is used to profile this network.
Definition: NvInfer.h:10567
Builds an engine from a network definition.
Definition: NvInfer.h:11219
int32_t getNbDLACores() const noexcept
Return the number of DLA engines available to this builder.
Definition: NvInfer.h:11238
IErrorRecorder * getErrorRecorder() const noexcept
get the ErrorRecorder assigned to this interface.
Definition: NvInfer.h:11345
apiv::VBuilder * mImpl
Definition: NvInfer.h:11477
ILogger * getLogger() const noexcept
get the logger with which the builder was created
Definition: NvInfer.h:11431
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:11421
int32_t getMaxThreads() const noexcept
get the maximum number of threads that can be used by the builder.
Definition: NvInfer.h:11461
IPluginRegistry & getPluginRegistry() noexcept
get the local plugin registry that can be used by the builder.
Definition: NvInfer.h:11471
nvinfer1::IOptimizationProfile * createOptimizationProfile() noexcept
Create a new optimization profile.
Definition: NvInfer.h:11311
void setGpuAllocator(IGpuAllocator *allocator) noexcept
Set the GPU allocator.
Definition: NvInfer.h:11256
nvinfer1::INetworkDefinition * createNetworkV2(NetworkDefinitionCreationFlags flags) noexcept
Create a network definition object.
Definition: NvInfer.h:11296
nvinfer1::IBuilderConfig * createBuilderConfig() noexcept
Create a builder configuration object.
Definition: NvInfer.h:11270
void reset() noexcept
Resets the builder state to default values.
Definition: NvInfer.h:11353
bool setMaxThreads(int32_t maxThreads) noexcept
Set the maximum number of threads.
Definition: NvInfer.h:11447
void setErrorRecorder(IErrorRecorder *recorder) noexcept
Set the ErrorRecorder for this interface.
Definition: NvInfer.h:11330
nvinfer1::IHostMemory * buildSerializedNetwork(INetworkDefinition &network, IBuilderConfig &config) noexcept
Builds and serializes a network for the given INetworkDefinition and IBuilderConfig.
Definition: NvInfer.h:11372
virtual ~IBuilder() noexcept=default
bool buildSerializedNetworkToStream(INetworkDefinition &network, IBuilderConfig &config, IStreamWriter &writer) noexcept
Builds and serializes a network into stream for the given INetworkDefinition and IBuilderConfig.
Definition: NvInfer.h:11393
A cast layer in a network.
Definition: NvInfer.h:3865
virtual ~ICastLayer() noexcept=default
apiv::VCastLayer * mImpl
Definition: NvInfer.h:3891
DataType getToType() const noexcept
Return cast layer output type.
Definition: NvInfer.h:3885
void setToType(DataType toType) noexcept
Set cast layer output type.
Definition: NvInfer.h:3874
A concatenation layer in a network definition.
Definition: NvInfer.h:2012
void setAxis(int32_t axis) noexcept
Set the axis along which concatenation occurs.
Definition: NvInfer.h:2025
int32_t getAxis() const noexcept
Get the axis along which concatenation occurs.
Definition: NvInfer.h:2035
virtual ~IConcatenationLayer() noexcept=default
This layer represents a condition input to an IIfConditional.
Definition: NvInfer.h:4528
virtual ~IConditionLayer() noexcept=default
Layer that represents a constant value.
Definition: NvInfer.h:3904
void setWeights(Weights weights) noexcept
Set the weights for the layer.
Definition: NvInfer.h:3914
Weights getWeights() const noexcept
Get the weights for the layer.
Definition: NvInfer.h:3924
void setDimensions(Dims const &dimensions) noexcept
Set the dimensions for the layer.
Definition: NvInfer.h:3936
apiv::VConstantLayer * mImpl
Definition: NvInfer.h:3954
virtual ~IConstantLayer() noexcept=default
Dims getDimensions() const noexcept
Get the dimensions for the layer.
Definition: NvInfer.h:3948
A convolution layer in a network definition.
Definition: NvInfer.h:982
Dims getPrePadding() const noexcept
Get the pre-padding.
Definition: NvInfer.h:1107
Weights getBiasWeights() const noexcept
Get the bias weights for the convolution.
Definition: NvInfer.h:1080
void setPaddingMode(PaddingMode paddingMode) noexcept
Set the padding mode.
Definition: NvInfer.h:1148
void setDilationNd(Dims const &dilation) noexcept
Set the multi-dimension dilation of the convolution.
Definition: NvInfer.h:1252
Dims getPaddingNd() const noexcept
Get the multi-dimension padding of the convolution.
Definition: NvInfer.h:1238
Dims getStrideNd() const noexcept
Get the multi-dimension stride of the convolution.
Definition: NvInfer.h:1208
Weights getKernelWeights() const noexcept
Get the kernel weights of the convolution.
Definition: NvInfer.h:1055
void setStrideNd(Dims const &stride) noexcept
Set the multi-dimension stride of the convolution.
Definition: NvInfer.h:1198
Dims getDilationNd() const noexcept
Get the multi-dimension dilation of the convolution.
Definition: NvInfer.h:1262
int64_t getNbOutputMaps() const noexcept
Get the number of output maps for the convolution.
Definition: NvInfer.h:1001
void setKernelWeights(Weights weights) noexcept
Set the kernel weights for the convolution.
Definition: NvInfer.h:1045
Dims getPostPadding() const noexcept
Get the post-padding.
Definition: NvInfer.h:1134
int64_t getNbGroups() const noexcept
Get the number of groups of the convolution.
Definition: NvInfer.h:1031
PaddingMode getPaddingMode() const noexcept
Get the padding mode.
Definition: NvInfer.h:1160
virtual ~IConvolutionLayer() noexcept=default
void setNbGroups(int64_t nbGroups) noexcept
Set the number of groups for a convolution.
Definition: NvInfer.h:1021
void setNbOutputMaps(int64_t nbOutputMaps) noexcept
Set the number of output maps for the convolution.
Definition: NvInfer.h:991
void setBiasWeights(Weights weights) noexcept
Set the bias weights for the convolution.
Definition: NvInfer.h:1070
Dims getKernelSizeNd() const noexcept
Get the multi-dimension kernel size of the convolution.
Definition: NvInfer.h:1183
void setPaddingNd(Dims const &padding) noexcept
Set the multi-dimension padding of the convolution.
Definition: NvInfer.h:1226
void setPrePadding(Dims const &padding) noexcept
Set the multi-dimension pre-padding of the convolution.
Definition: NvInfer.h:1097
void setPostPadding(Dims const &padding) noexcept
Set the multi-dimension post-padding of the convolution.
Definition: NvInfer.h:1124
void setKernelSizeNd(Dims const &kernelSize) noexcept
Set the multi-dimension kernel size of the convolution.
Definition: NvInfer.h:1173
Layer that represents a cumulative operation across a tensor.
Definition: NvInfer.h:6719
bool setOperation(CumulativeOperation op) noexcept
Set the cumulative operation for the layer.
Definition: NvInfer.h:6730
void setReverse(bool reverse) noexcept
Specify whether the cumulative operation should be applied backward.
Definition: NvInfer.h:6778
apiv::VCumulativeLayer * mImpl
Definition: NvInfer.h:6796
bool getExclusive() const noexcept
Get whether it is exclusive accumulation or inclusive accumulation.
Definition: NvInfer.h:6766
virtual ~ICumulativeLayer() noexcept=default
bool getReverse() const noexcept
Get the boolean that specifies whether the cumulative operation should be applied backward.
Definition: NvInfer.h:6790
void setExclusive(bool exclusive) noexcept
Set whether it is an exclusive accumulation or inclusive accumulation.
Definition: NvInfer.h:6754
CumulativeOperation getOperation() const noexcept
Get the cumulative operation for the layer.
Definition: NvInfer.h:6742
A deconvolution layer in a network definition.
Definition: NvInfer.h:2053
void setBiasWeights(Weights weights) noexcept
Set the bias weights for the deconvolution.
Definition: NvInfer.h:2141
int64_t getNbGroups() const noexcept
Get the number of groups for a deconvolution.
Definition: NvInfer.h:2102
Weights getKernelWeights() const noexcept
Get the kernel weights for the deconvolution.
Definition: NvInfer.h:2126
void setPrePadding(Dims const &padding) noexcept
Set the multi-dimension pre-padding of the deconvolution.
Definition: NvInfer.h:2168
Dims getStrideNd() const noexcept
Get the multi-dimension stride of the deconvolution.
Definition: NvInfer.h:2283
Dims getDilationNd() const noexcept
Get the multi-dimension dilation of the deconvolution.
Definition: NvInfer.h:2349
Weights getBiasWeights() const noexcept
Get the bias weights for the deconvolution.
Definition: NvInfer.h:2151
void setKernelWeights(Weights weights) noexcept
Set the kernel weights for the deconvolution.
Definition: NvInfer.h:2116
int64_t getNbOutputMaps() const noexcept
Get the number of output feature maps for the deconvolution.
Definition: NvInfer.h:2072
void setStrideNd(Dims const &stride) noexcept
Set the multi-dimension stride of the deconvolution.
Definition: NvInfer.h:2273
Dims getPostPadding() const noexcept
Get the padding.
Definition: NvInfer.h:2205
Dims getKernelSizeNd() const noexcept
Get the multi-dimension kernel size of the deconvolution.
Definition: NvInfer.h:2256
void setPostPadding(Dims const &padding) noexcept
Set the multi-dimension post-padding of the deconvolution.
Definition: NvInfer.h:2195
void setKernelSizeNd(Dims const &kernelSize) noexcept
Set the multi-dimension kernel size of the deconvolution.
Definition: NvInfer.h:2246
virtual ~IDeconvolutionLayer() noexcept=default
void setPaddingNd(Dims const &padding) noexcept
Set the multi-dimension padding of the deconvolution.
Definition: NvInfer.h:2301
void setNbOutputMaps(int64_t nbOutputMaps) noexcept
Set the number of output feature maps for the deconvolution.
Definition: NvInfer.h:2062
Dims getPaddingNd() const noexcept
Get the multi-dimension padding of the deconvolution.
Definition: NvInfer.h:2313
void setDilationNd(Dims const &dilation) noexcept
Set the multi-dimension dilation of the deconvolution.
Definition: NvInfer.h:2339
void setPaddingMode(PaddingMode paddingMode) noexcept
Set the padding mode.
Definition: NvInfer.h:2219
void setNbGroups(int64_t nbGroups) noexcept
Set the number of groups for a deconvolution.
Definition: NvInfer.h:2092
Dims getPrePadding() const noexcept
Get the pre-padding.
Definition: NvInfer.h:2178
PaddingMode getPaddingMode() const noexcept
Get the padding mode.
Definition: NvInfer.h:2231
A Dequantize layer in a network definition.
Definition: NvInfer.h:5617
TRT_NODISCARD Dims getBlockShape() const noexcept
Get the shape of the quantization block.
Definition: NvInfer.h:5666
void setToType(DataType toType) noexcept
Set the Dequantize layer output type.
Definition: NvInfer.h:5682
virtual ~IDequantizeLayer() noexcept=default
int32_t getAxis() const noexcept
Get the quantization axis.
Definition: NvInfer.h:5627
bool setBlockShape(Dims const &blockShape) noexcept
Set the shape of the quantization block.
Definition: NvInfer.h:5655
DataType getToType() const noexcept
Return the Dequantize layer output type.
Definition: NvInfer.h:5694
void setAxis(int32_t axis) noexcept
Set the quantization axis.
Definition: NvInfer.h:5638
Definition: NvInfer.h:7884
virtual ~IDistCollectiveLayer() noexcept=default
A network layer to perform dynamic quantization.
Definition: NvInfer.h:5722
DataType getScaleType() const noexcept
Return the scale factors data type.
Definition: NvInfer.h:5788
TRT_DEPRECATED void setAxis(int32_t axis) noexcept
Set the axis along which block quantization occurs.
Definition: NvInfer.h:5801
TRT_DEPRECATED void setBlockSize(int32_t size) noexcept
Set the size of the quantization block.
Definition: NvInfer.h:5824
Dims getBlockShape() const noexcept
Get the shape of the quantization block.
Definition: NvInfer.h:5859
void setScaleType(DataType scaleType) noexcept
Set the data type of the scale factors used to quantize the data.
Definition: NvInfer.h:5775
DataType getToType() const noexcept
Return DynamicQuantizeLayer's quantized output type.
Definition: NvInfer.h:5762
TRT_DEPRECATED int32_t getAxis() const noexcept
Get the axis along which blocking occurs.
Definition: NvInfer.h:5811
virtual ~IDynamicQuantizeLayer() noexcept=default
void setToType(DataType toType) noexcept
Set DynamicQuantizeLayer's quantized output type.
Definition: NvInfer.h:5749
void setBlockShape(Dims const &blockShape) noexcept
Set the shape of the quantization block.
Definition: NvInfer.h:5847
TRT_DEPRECATED int32_t getBlockSize() const noexcept
Get the size of the quantization block.
Definition: NvInfer.h:5834
An Einsum layer in a network.
Definition: NvInfer.h:5904
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:5915
virtual ~IEinsumLayer() noexcept=default
char const * getEquation() const noexcept
Return the equation.
Definition: NvInfer.h:5925
A elementwise layer in a network definition.
Definition: NvInfer.h:2423
virtual ~IElementWiseLayer() noexcept=default
apiv::VElementWiseLayer * mImpl
Definition: NvInfer.h:2452
ElementWiseOperation getOperation() const noexcept
Get the binary operation for the layer.
Definition: NvInfer.h:2446
void setOperation(ElementWiseOperation op) noexcept
Set the binary operation for the layer.
Definition: NvInfer.h:2434
Generate a tensor according to a specified mode.
Definition: NvInfer.h:5118
bool isAlphaBetaInt64() const noexcept
Return true if alpha/beta have type int64, false if they have type double.
Definition: NvInfer.h:5350
FillOperation getOperation() const noexcept
Get the fill operation for the layer.
Definition: NvInfer.h:5164
void setOperation(FillOperation op) noexcept
Set the fill operation for the layer.
Definition: NvInfer.h:5154
DataType getToType() const noexcept
Get the fill layer output type.
Definition: NvInfer.h:5380
void setAlphaInt64(int64_t alpha) noexcept
Set the alpha parameter with int64 datatype.
Definition: NvInfer.h:5293
void setBetaInt64(int64_t beta) noexcept
Set the beta parameter with int64 datatype.
Definition: NvInfer.h:5327
void setBeta(double beta) noexcept
Set the beta parameter.
Definition: NvInfer.h:5217
int64_t getAlphaInt64() const noexcept
Get the value of alpha parameter with int64 datatype.
Definition: NvInfer.h:5308
int64_t getBetaInt64() const noexcept
Get the value of beta parameter with int64 datatype.
Definition: NvInfer.h:5342
double getAlpha() const noexcept
Get the value of alpha parameter.
Definition: NvInfer.h:5198
void setDimensions(Dims const &dimensions) noexcept
Set the output tensor's dimensions.
Definition: NvInfer.h:5129
void setAlpha(double alpha) noexcept
Set the alpha parameter.
Definition: NvInfer.h:5183
void setToType(DataType toType) noexcept
Set the fill layer output type.
Definition: NvInfer.h:5368
Dims getDimensions() const noexcept
Get the output tensor's dimensions.
Definition: NvInfer.h:5144
double getBeta() const noexcept
Get the value of beta parameter.
Definition: NvInfer.h:5232
virtual ~IFillLayer() noexcept=default
A Gather layer in a network definition. Supports several kinds of gathering.
Definition: NvInfer.h:2556
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:2567
void setNbElementWiseDims(int32_t elementWiseDims) noexcept
Set the number of leading dimensions of indices tensor to be handled elementwise.
Definition: NvInfer.h:2602
apiv::VGatherLayer * mImpl
Definition: NvInfer.h:2638
int32_t getNbElementWiseDims() const noexcept
Get the number of leading dimensions of indices tensor to be handled elementwise.
Definition: NvInfer.h:2612
void setMode(GatherMode mode) noexcept
Set the gather mode.
Definition: NvInfer.h:2622
int32_t getGatherAxis() const noexcept
Get the axis to gather on.
Definition: NvInfer.h:2579
GatherMode getMode() const noexcept
Get the gather mode.
Definition: NvInfer.h:2632
virtual ~IGatherLayer() noexcept=default
A GridSample layer in a network definition.
Definition: NvInfer.h:6126
void setInterpolationMode(InterpolationMode mode) noexcept
Set the grid sample interpolation mode.
Definition: NvInfer.h:6133
bool setSampleMode(SampleMode mode) noexcept
Set the sample mode.
Definition: NvInfer.h:6179
void setAlignCorners(bool alignCorners) noexcept
Set the align corners mode.
Definition: NvInfer.h:6155
apiv::VGridSampleLayer * mImpl
Definition: NvInfer.h:6197
SampleMode getSampleMode() const noexcept
Get the sample mode.
Definition: NvInfer.h:6191
InterpolationMode getInterpolationMode() const noexcept
Get the grid sample interpolation mode.
Definition: NvInfer.h:6145
bool getAlignCorners() const noexcept
Get the align corners mode.
Definition: NvInfer.h:6167
virtual ~IGridSampleLayer() noexcept=default
Class to handle library allocated memory that is accessible to the user.
Definition: NvInferRuntime.h:142
A layer that represents the identity function.
Definition: NvInfer.h:3852
apiv::VIdentityLayer * mImpl
Definition: NvInfer.h:3854
virtual ~IIdentityLayer() noexcept=default
This is a base class for Conditional boundary layers.
Definition: NvInfer.h:4507
IIfConditional * getConditional() const noexcept
Get a pointer to the IIfConditional associated with this boundary layer.
Definition: NvInfer.h:4512
virtual ~IIfConditionalBoundaryLayer() noexcept=default
Helper for constructing conditionally-executed subgraphs.
Definition: NvInfer.h:4590
IIfConditionalInputLayer * addInput(ITensor &input) noexcept
Add an If-conditional input.
Definition: NvInfer.h:4631
char const * getName() const noexcept
Return the name of the conditional.
Definition: NvInfer.h:4656
virtual ~IIfConditional() noexcept=default
IConditionLayer * setCondition(ITensor &condition) noexcept
Set the condition tensor for this If-Conditional construct.
Definition: NvInfer.h:4601
IIfConditionalOutputLayer * addOutput(ITensor &trueSubgraphOutput, ITensor &falseSubgraphOutput) noexcept
Add an If-conditional output.
Definition: NvInfer.h:4619
void setName(char const *name) noexcept
Set the name of the conditional.
Definition: NvInfer.h:4646
This layer represents an output of an IIfConditional.
Definition: NvInfer.h:4545
virtual ~IIfConditionalOutputLayer() noexcept=default
A layer to do iterations.
Definition: NvInfer.h:4821
virtual ~IIteratorLayer() noexcept=default
void setReverse(bool reverse) noexcept
Set iteration order to be reverse.
Definition: NvInfer.h:4848
bool getReverse() const noexcept
Check if the iteration order is reverse.
Definition: NvInfer.h:4858
int32_t getAxis() const noexcept
Get axis being iterated over.
Definition: NvInfer.h:4834
void setAxis(int32_t axis) noexcept
Set axis to iterate over.
Definition: NvInfer.h:4826
Layer that represents a KVCacheUpdate operation.
Definition: NvInfer.h:7386
bool setCacheMode(KVCacheMode cacheMode) noexcept
Set the mode of the KVCacheUpdate layer.
Definition: NvInfer.h:7409
virtual ~IKVCacheUpdateLayer() noexcept=default
KVCacheMode getCacheMode() const noexcept
Get the mode of the KVCacheUpdate layer.
Definition: NvInfer.h:7419
apiv::VKVCacheUpdateLayer * mImpl
Definition: NvInfer.h:7425
A LRN layer in a network definition.
Definition: NvInfer.h:1667
int64_t getWindowSize() const noexcept
Get the LRN window size.
Definition: NvInfer.h:1688
float getAlpha() const noexcept
Get the LRN alpha value.
Definition: NvInfer.h:1710
void setWindowSize(int64_t windowSize) noexcept
Set the LRN window size.
Definition: NvInfer.h:1678
void setK(float k) noexcept
Set the LRN K value.
Definition: NvInfer.h:1744
void setAlpha(float alpha) noexcept
Set the LRN alpha value.
Definition: NvInfer.h:1700
void setBeta(float beta) noexcept
Set the LRN beta value.
Definition: NvInfer.h:1722
virtual ~ILRNLayer() noexcept=default
float getBeta() const noexcept
Get the LRN beta value.
Definition: NvInfer.h:1732
float getK() const noexcept
Get the LRN K value.
Definition: NvInfer.h:1754
Base class for all layer classes in a network definition.
Definition: NvInfer.h:464
TRT_DEPRECATED void setPrecision(DataType dataType) noexcept
Set the preferred or required computational precision of this layer in a weakly-typed network.
Definition: NvInfer.h:584
TRT_DEPRECATED void setOutputType(int32_t index, DataType dataType) noexcept
Set the output type of this layer in a weakly-typed network.
Definition: NvInfer.h:672
TRT_DEPRECATED bool precisionIsSet() const noexcept
whether the computational precision has been set for this layer
Definition: NvInfer.h:610
void setMetadata(char const *metadata) noexcept
Set the metadata for this layer.
Definition: NvInfer.h:735
TRT_DEPRECATED void resetOutputType(int32_t index) noexcept
reset the output type for this layer
Definition: NvInfer.h:717
void setName(char const *name) noexcept
Set the name of a layer.
Definition: NvInfer.h:485
int32_t getNbInputs() const noexcept
Get the number of inputs of a layer.
Definition: NvInfer.h:503
int32_t getNbRanks() const noexcept
Get the number of ranks for multi-device execution.
Definition: NvInfer.h:781
char const * getMetadata() const noexcept
Get the metadata of the layer.
Definition: NvInfer.h:748
DataType getOutputType(int32_t index) const noexcept
get the output type of this layer
Definition: NvInfer.h:687
DataType getPrecision() const noexcept
get the computational precision of this layer
Definition: NvInfer.h:596
TRT_DEPRECATED bool outputTypeIsSet(int32_t index) const noexcept
whether the output type has been set for this layer
Definition: NvInfer.h:703
char const * getName() const noexcept
Return the name of a layer.
Definition: NvInfer.h:495
int32_t getNbOutputs() const noexcept
Get the number of outputs of a layer.
Definition: NvInfer.h:524
ITensor * getOutput(int32_t index) const noexcept
Get the layer output corresponding to the given index.
Definition: NvInfer.h:534
void setInput(int32_t index, ITensor &tensor) noexcept
Replace an input of this layer with a specific tensor.
Definition: NvInfer.h:551
ITensor * getInput(int32_t index) const noexcept
Get the layer input corresponding to the given index.
Definition: NvInfer.h:516
bool setNbRanks(int32_t nbRanks) noexcept
Set the number of ranks for multi-device execution.
Definition: NvInfer.h:769
LayerType getType() const noexcept
Return the type of a layer.
Definition: NvInfer.h:471
TRT_DEPRECATED void resetPrecision() noexcept
reset the computational precision for this layer
Definition: NvInfer.h:622
virtual ~ILayer() noexcept=default
Application-implemented logging interface for the builder, refitter and runtime.
Definition: NvInferRuntime.h:1588
This is a base class for Loop boundary layers.
Definition: NvInfer.h:4484
virtual ~ILoopBoundaryLayer() noexcept=default
ILoop * getLoop() const noexcept
Get a pointer to ILoop associated with this boundary layer.
Definition: NvInfer.h:4489
Helper for creating a recurrent subgraph.
Definition: NvInfer.h:4879
void setName(char const *name) noexcept
Set the name of the loop.
Definition: NvInfer.h:4949
ITripLimitLayer * addTripLimit(ITensor &tensor, TripLimit limit) noexcept
Add a trip-count limiter, based on the given tensor.
Definition: NvInfer.h:4908
IIteratorLayer * addIterator(ITensor &tensor, int32_t axis=0, bool reverse=false) noexcept
Return layer that subscripts tensor by loop iteration.
Definition: NvInfer.h:4921
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:4934
virtual ~ILoop() noexcept=default
char const * getName() const noexcept
Return the name of the loop.
Definition: NvInfer.h:4959
IRecurrenceLayer * addRecurrence(ITensor &initialValue) noexcept
Create a recurrence layer for this loop with initialValue as its first input.
Definition: NvInfer.h:4887
An ILoopOutputLayer is the sole way to get output from a loop.
Definition: NvInfer.h:4721
virtual ~ILoopOutputLayer() noexcept=default
int32_t getAxis() const noexcept
Get axis being concatenated over.
Definition: NvInfer.h:4751
LoopOutput getLoopOutput() const noexcept
Get which kind a loop output has.
Definition: NvInfer.h:4726
void setAxis(int32_t axis) noexcept
Set where to insert the contenation axis. Ignored if getLoopOutput() is kLAST_VALUE.
Definition: NvInfer.h:4743
Layer that represents a Matrix Multiplication.
Definition: NvInfer.h:3699
apiv::VMatrixMultiplyLayer * mImpl
Definition: NvInfer.h:3727
virtual ~IMatrixMultiplyLayer() noexcept=default
MatrixOperation getOperation(int32_t index) const noexcept
Get the operation for an input tensor.
Definition: NvInfer.h:3721
void setOperation(int32_t index, MatrixOperation op) noexcept
Set the operation for an input tensor.
Definition: NvInfer.h:3709
A MoE layer in a network definition. Mixture of Experts (MoE) is a collection of experts with each ex...
Definition: NvInfer.h:7568
void setSwigluParamLimit(float limit) noexcept
Set the SwiGLU parameter limit.
Definition: NvInfer.h:7783
void setDynQOutputScaleType(DataType type) noexcept
Set the dynamic quantization output scale type.
Definition: NvInfer.h:7736
MoEActType getActivationType() const noexcept
Get the activation type for the MoE layer.
Definition: NvInfer.h:7617
void setQuantizationToType(DataType type) noexcept
Set the data type the mul output is quantized to.
Definition: NvInfer.h:7684
void setQuantizationDynamicDblQ(ITensor &fcDownActivationDblQScale, DataType dataType, Dims const &blockShape, DataType dynQOutputScaleType) noexcept
Configure dynamic quantization (with double quantization) after the mul op. ┌── fcGate ── activation ...
Definition: NvInfer.h:7669
void setQuantizationStatic(ITensor &fcDownActivationScale, DataType dataType) noexcept
Configure static quantization after the mul op. ┌── fcGate ── activation ───┐ │ │ hiddenStates ───┤ ├...
Definition: NvInfer.h:7640
virtual ~IMoELayer() noexcept=default
float getSwigluParamLimit() const noexcept
Get the SwiGLU parameter limit.
Definition: NvInfer.h:7795
DataType getQuantizationToType() const noexcept
Get the data type the mul in MoE layer is quantized to.
Definition: NvInfer.h:7696
DataType getDynQOutputScaleType() const noexcept
Get the dynamic quantization output scale type.
Definition: NvInfer.h:7748
void setActivationType(MoEActType activationType) noexcept
Set the activation type for the MoE layer.
Definition: NvInfer.h:7605
Dims getQuantizationBlockShape() const noexcept
Get the block shape for the quantization of the Mul output.
Definition: NvInfer.h:7724
void setGatedWeights(ITensor &fcGateWeights, ITensor &fcUpWeights, ITensor &fcDownWeights, MoEActType activationType) noexcept
Set the weights of the experts when each expert is a GLU (gated linear unit). In each GLU,...
Definition: NvInfer.h:7581
float getSwigluParamBeta() const noexcept
Get the SwiGLU parameter beta.
Definition: NvInfer.h:7847
void setSwigluParamBeta(float beta) noexcept
Set the SwiGLU parameter beta.
Definition: NvInfer.h:7835
void setGatedBiases(ITensor &fcGateBiases, ITensor &fcUpBiases, ITensor &fcDownBiases) noexcept
Set the biases of the experts when each expert is a GLU (gated linear unit). In each GLU,...
Definition: NvInfer.h:7593
void setSwigluParams(float limit, float alpha, float beta) noexcept
Set the SwiGLU parameters.
Definition: NvInfer.h:7769
void setQuantizationBlockShape(Dims const &blockShape) noexcept
Set the block shape for the quantization of the Mul output.
Definition: NvInfer.h:7712
void setInput(int32_t index, ITensor &tensor) noexcept
Set the input of the MoE layer.
Definition: NvInfer.h:7864
float getSwigluParamAlpha() const noexcept
Get the SwiGLU parameter alpha.
Definition: NvInfer.h:7821
void setSwigluParamAlpha(float alpha) noexcept
Set the SwiGLU parameter alpha.
Definition: NvInfer.h:7809
A non-maximum suppression layer in a network definition.
Definition: NvInfer.h:6278
virtual ~INMSLayer() noexcept=default
void setTopKBoxLimit(int32_t limit) noexcept
Set the TopK box limit parameter for the layer.
Definition: NvInfer.h:6315
void setBoundingBoxFormat(BoundingBoxFormat fmt) noexcept
Set the bounding box format parameter for the layer.
Definition: NvInfer.h:6289
BoundingBoxFormat getBoundingBoxFormat() const noexcept
Get the bounding box format parameter for the layer.
Definition: NvInfer.h:6301
bool setIndicesType(DataType type) noexcept
Set the indices type for the layer.
Definition: NvInfer.h:6360
apiv::VNMSLayer * mImpl
Definition: NvInfer.h:6378
int32_t getTopKBoxLimit() const noexcept
Get the TopK box limit parameter for the layer.
Definition: NvInfer.h:6325
DataType getIndicesType() const noexcept
Return the NMS layer indices type.
Definition: NvInfer.h:6372
A network definition for input to the builder.
Definition: NvInfer.h:7908
IConcatenationLayer * addConcatenation(ITensor *const *inputs, int32_t nbInputs) noexcept
Add a concatenation layer to the network.
Definition: NvInfer.h:8136
IShuffleLayer * addShuffle(ITensor &input) noexcept
Add a shuffle layer to the network.
Definition: NvInfer.h:8199
void setName(char const *name) noexcept
Sets the name of the network.
Definition: NvInfer.h:8625
ITopKLayer * addTopK(ITensor &input, TopKOperation op, int32_t k, uint32_t reduceAxes, DataType indicesType) noexcept
Add a TopK layer to the network.
Definition: NvInfer.h:8395
bool markDebug(ITensor &tensor) noexcept
Mark a tensor as a debug tensor.
Definition: NvInfer.h:7979
ILRNLayer * addLRN(ITensor &input, int64_t window, float alpha, float beta, float k) noexcept
Add a LRN layer to the network.
Definition: NvInfer.h:8080
ICumulativeLayer * addCumulative(ITensor &input, ITensor &axis, CumulativeOperation operation, bool exclusive, bool reverse) noexcept
Add a cumulative layer to the network.
Definition: NvInfer.h:9292
IAssertionLayer * addAssertion(ITensor &condition, char const *message) noexcept
Add an assertion layer to the network.
Definition: NvInfer.h:8927
TRT_DEPRECATED INonZeroLayer * addNonZero(ITensor &input) noexcept
Add a nonzero layer to the network.
Definition: NvInfer.h:8486
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:8746
ICastLayer * addCast(ITensor &input, DataType toType) noexcept
Add a cast layer.
Definition: NvInfer.h:8555
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:8825
char const * getName() const noexcept
Returns the name associated with the network.
Definition: NvInfer.h:8639
IParametricReLULayer * addParametricReLU(ITensor &input, ITensor &slope) noexcept
Add a parametric ReLU layer to the network.
Definition: NvInfer.h:8724
ITensor * getOutput(int32_t index) const noexcept
Get the output tensor specified by the given index.
Definition: NvInfer.h:8300
ITensor * getInput(int32_t index) const noexcept
Get the input tensor specified by the given index.
Definition: NvInfer.h:8270
TRT_DEPRECATED ITopKLayer * addTopK(ITensor &input, TopKOperation op, int32_t k, uint32_t reduceAxes) noexcept
Add a TopK layer to the network.
Definition: NvInfer.h:8362
IDequantizeLayer * addDequantize(ITensor &input, ITensor &scale, DataType outputType) noexcept
Add a dequantization layer to the network.
Definition: NvInfer.h:9050
bool unmarkOutputForShapes(ITensor &tensor) noexcept
Undo markOutputForShapes.
Definition: NvInfer.h:8706
IFillLayer * addFill(Dims const &dimensions, FillOperation op, DataType outputType) noexcept
Add a fill layer to the network.
Definition: NvInfer.h:8953
ILoop * addLoop() noexcept
Add a loop to the network.
Definition: NvInfer.h:8856
bool markUnfusedTensorsAsDebugTensors() noexcept
Mark unfused tensors as debug tensors.
Definition: NvInfer.h:8027
TRT_NODISCARD INormalizationLayer * addNormalizationV2(ITensor &input, ITensor &scale, ITensor &bias, uint32_t axesMask) noexcept
Add a normalization layer to the network.
Definition: NvInfer.h:9542
IActivationLayer * addActivation(ITensor &input, ActivationType type) noexcept
Add an activation layer to the network.
Definition: NvInfer.h:8061
ISliceLayer * addSlice(ITensor &input, Dims const &start, Dims const &size, Dims const &stride) noexcept
Add a slice layer to the network.
Definition: NvInfer.h:8601
virtual ~INetworkDefinition() noexcept=default
virtual IBuilder & getBuilder() const noexcept
Return the builder from which this INetworkDefinition was created.
Definition: NvInfer.h:9438
ILayer * getLayer(int32_t index) const noexcept
Get the layer specified by the given index.
Definition: NvInfer.h:8242
bool isDebugTensor(ITensor const &tensor) const noexcept
Check if a tensor is marked as debug tensor.
Definition: NvInfer.h:8005
bool getFlag(NetworkDefinitionCreationFlag networkDefinitionCreationFlag) const noexcept
Returns true if the network definition creation flag is set.
Definition: NvInfer.h:8677
IIfConditional * addIfConditional() noexcept
Add an if-then-else to the network.
Definition: NvInfer.h:8871
IErrorRecorder * getErrorRecorder() const noexcept
get the ErrorRecorder assigned to this interface.
Definition: NvInfer.h:9027
ISqueezeLayer * addSqueeze(ITensor &input, ITensor &axes) noexcept
Add a squeeze layer to the network.
Definition: NvInfer.h:9495
TRT_DEPRECATED INMSLayer * addNMS(ITensor &boxes, ITensor &scores, ITensor &maxOutputBoxesPerClass) noexcept
Add a non-maximum suppression layer to the network.
Definition: NvInfer.h:9201
IReverseSequenceLayer * addReverseSequence(ITensor &input, ITensor &sequenceLens) noexcept
Add a ReverseSequence layer to the network.
Definition: NvInfer.h:9238
TRT_DEPRECATED IDynamicQuantizeLayer * addDynamicQuantize(ITensor &input, int32_t axis, int32_t blockSize, DataType outputType, DataType scaleType) noexcept
Add a dynamic quantization layer to the network.
Definition: NvInfer.h:9121
int32_t getNbInputs() const noexcept
Get the number of inputs in the network.
Definition: NvInfer.h:8254
NetworkDefinitionCreationFlags getFlags() const noexcept
Get the network definition creation flags for this network definition object. Defaults to 0.
Definition: NvInfer.h:8665
IQuantizeLayer * addQuantize(ITensor &input, ITensor &scale, DataType outputType) noexcept
Add a quantization layer to the network.
Definition: NvInfer.h:9094
IDynamicQuantizeLayer * addDynamicQuantizeV2(ITensor &input, Dims const &blockShape, DataType outputType, DataType scaleType) noexcept
Add a dynamic quantization layer to the network.
Definition: NvInfer.h:9145
IReduceLayer * addReduce(ITensor &input, ReduceOperation operation, uint32_t reduceAxes, bool keepDimensions) noexcept
Add a reduce layer to the network.
Definition: NvInfer.h:8326
IUnaryLayer * addUnary(ITensor &input, UnaryOperation operation) noexcept
Add a unary layer to the network.
Definition: NvInfer.h:8185
IGridSampleLayer * addGridSample(ITensor &input, ITensor &grid) noexcept
Add a GridSample layer to the network.
Definition: NvInfer.h:9179
void removeTensor(ITensor &tensor) noexcept
remove a tensor from the network definition.
Definition: NvInfer.h:8570
bool areWeightsMarkedRefittable(char const *name) const noexcept
Whether the weight has been marked as refittable.
Definition: NvInfer.h:9476
ISelectLayer * addSelect(ITensor &condition, ITensor &thenInput, ITensor &elseInput) noexcept
Add a select layer to the network.
Definition: NvInfer.h:8910
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:9070
TRT_DEPRECATED INormalizationLayer * addNormalization(ITensor &input, ITensor &scale, ITensor &bias, uint32_t axesMask) noexcept
Add a normalization layer to the network.
Definition: NvInfer.h:9270
int32_t getNbLayers() const noexcept
Get the number of layers in the network.
Definition: NvInfer.h:8228
apiv::VNetworkDefinition * mImpl
Definition: NvInfer.h:9548
IKVCacheUpdateLayer * addKVCacheUpdate(ITensor &cache, ITensor &update, ITensor &writeIndices, KVCacheMode cacheMode) noexcept
Add a KVCacheUpdate layer to the network.
Definition: NvInfer.h:9378
bool markOutputForShapes(ITensor &tensor) noexcept
Enable tensor's value to be computed by IExecutionContext::getShapeBinding.
Definition: NvInfer.h:8694
IOneHotLayer * addOneHot(ITensor &indices, ITensor &values, ITensor &depth, int32_t axis) noexcept
Add a OneHot layer to the network.
Definition: NvInfer.h:8216
IScaleLayer * addScale(ITensor &input, ScaleMode mode, Weights shift, Weights scale, Weights power) noexcept
Add a Scale layer to the network.
Definition: NvInfer.h:8106
void unmarkOutput(ITensor &tensor) noexcept
unmark a tensor as a network output.
Definition: NvInfer.h:8582
IIdentityLayer * addIdentity(ITensor &input) noexcept
Add an identity layer.
Definition: NvInfer.h:8540
IGatherLayer * addGatherV2(ITensor &data, ITensor &indices, GatherMode mode) noexcept
Add gather with specified mode, axis=0 and nbElementWiseDims=0.
Definition: NvInfer.h:8427
INonZeroLayer * addNonZero(ITensor &input, DataType indicesType) noexcept
Add a nonzero layer to the network.
Definition: NvInfer.h:8502
IElementWiseLayer * addElementWise(ITensor &input1, ITensor &input2, ElementWiseOperation op) noexcept
Add an elementwise layer to the network.
Definition: NvInfer.h:8163
IConstantLayer * addConstant(Dims const &dimensions, Weights weights) noexcept
Add a constant layer to the network.
Definition: NvInfer.h:8526
void setErrorRecorder(IErrorRecorder *recorder) noexcept
Set the ErrorRecorder for this interface.
Definition: NvInfer.h:9012
IPoolingLayer * addPoolingNd(ITensor &input, PoolingType type, Dims const &windowSize) noexcept
Add a multi-dimension pooling layer to the network.
Definition: NvInfer.h:8766
INMSLayer * addNMS(ITensor &boxes, ITensor &scores, ITensor &maxOutputBoxesPerClass, DataType indicesType) noexcept
Add a non-maximum suppression layer to the network.
Definition: NvInfer.h:9221
IRaggedSoftMaxLayer * addRaggedSoftMax(ITensor &input, ITensor &bounds) noexcept
Add a RaggedSoftMax layer to the network.
Definition: NvInfer.h:8446
IShapeLayer * addShape(ITensor &input) noexcept
Add a shape layer to the network.
Definition: NvInfer.h:8655
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:8411
IAttention * addAttention(ITensor &query, ITensor &key, ITensor &value, AttentionNormalizationOp normOp, bool causal) noexcept
Add an attention to the network.
Definition: NvInfer.h:9319
bool unmarkWeightsRefittable(char const *name) noexcept
Unmark weights as refittable when the builder flag kREFIT_INDIVIDUAL is set.
Definition: NvInfer.h:9463
bool markWeightsRefittable(char const *name) noexcept
Mark weights as refittable when the builder flag kREFIT_INDIVIDUAL is set.
Definition: NvInfer.h:9451
IRotaryEmbeddingLayer * addRotaryEmbedding(ITensor &input, ITensor &cosCache, ITensor &sinCache, bool interleaved, int32_t rotaryEmbeddingDim) noexcept
Add a Rotary Position Embedding (RoPE) layer to the network.
Definition: NvInfer.h:9344
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:8788
IResizeLayer * addResize(ITensor &input) noexcept
Add a resize layer to the network.
Definition: NvInfer.h:8842
IUnsqueezeLayer * addUnsqueeze(ITensor &input, ITensor &axes) noexcept
Add an unsqueeze layer to the network.
Definition: NvInfer.h:9516
IMatrixMultiplyLayer * addMatrixMultiply(ITensor &input0, MatrixOperation op0, ITensor &input1, MatrixOperation op1) noexcept
Add a MatrixMultiply layer to the network.
Definition: NvInfer.h:8467
ISoftMaxLayer * addSoftMax(ITensor &input) noexcept
Add a SoftMax layer to the network.
Definition: NvInfer.h:8119
bool unmarkDebug(ITensor &tensor) noexcept
Unmark a tensor as a debug tensor.
Definition: NvInfer.h:7995
IEinsumLayer * addEinsum(ITensor *const *inputs, int32_t nbInputs, char const *equation) noexcept
Add an Einsum layer to the network.
Definition: NvInfer.h:9161
void markOutput(ITensor &tensor) noexcept
Mark a tensor as a network output.
Definition: NvInfer.h:7961
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:8969
int32_t getNbOutputs() const noexcept
Get the number of outputs in the network.
Definition: NvInfer.h:8284
bool setWeightsName(Weights weights, char const *name) noexcept
Associate a name with all current uses of the given weights.
Definition: NvInfer.h:8993
TRT_NODISCARD IDistCollectiveLayer * addDistCollective(ITensor &input, CollectiveOperation distCollectiveOp, ReduceOperation reduceOp, int64_t root, int64_t *groups, int64_t groupSize) noexcept
Add a DistCollective layer to the network.
Definition: NvInfer.h:9426
IMoELayer * addMoE(ITensor &hiddenStates, ITensor &selectedExpertsForTokens, ITensor &scoresForSelectedExperts) noexcept
Add a MoE (Mixture of Experts) layer to the network.
Definition: NvInfer.h:9398
bool unmarkUnfusedTensorsAsDebugTensors() noexcept
Undo the marking of unfused tensors as debug tensors.
Definition: NvInfer.h:8041
Forward declaration of IEngineInspector for use by other interfaces.
Definition: NvInferRuntime.h:51
Definition: NvInfer.h:3753
DataType getIndicesType() const noexcept
Return the NonZero layer indices type.
Definition: NvInfer.h:3777
bool setIndicesType(DataType type) noexcept
Set the indices type for the layer.
Definition: NvInfer.h:3765
virtual ~INonZeroLayer() noexcept=default
A normalization layer in a network definition.
Definition: NvInfer.h:6467
float getEpsilon() const noexcept
Get the epsilon value used for the normalization calculation.
Definition: NvInfer.h:6486
TRT_DEPRECATED void setComputePrecision(DataType type) noexcept
Set the compute precision of this layer.
Definition: NvInfer.h:6565
uint32_t getAxes() const noexcept
Get the axes value used for the normalization calculation.
Definition: NvInfer.h:6506
virtual ~INormalizationLayer() noexcept=default
void setEpsilon(float eps) noexcept
Set the epsilon value used for the normalization calculation.
Definition: NvInfer.h:6476
TRT_NODISCARD bool isV2() const noexcept
Returns true if this layer was created through addNormalizationV2().
Definition: NvInfer.h:6587
apiv::VNormalizationLayer * mImpl
Definition: NvInfer.h:6593
int64_t getNbGroups() const noexcept
Get the number of groups used to split the channels for the normalization calculation.
Definition: NvInfer.h:6537
void setAxes(uint32_t axesMask) noexcept
Set the reduction axes for the normalization calculation.
Definition: NvInfer.h:6496
void setNbGroups(int64_t nbGroups) noexcept
Set the number of groups used to split the channels in the normalization calculation.
Definition: NvInfer.h:6527
TRT_DEPRECATED DataType getComputePrecision() const noexcept
Get the compute precision of this layer.
Definition: NvInfer.h:6577
A OneHot layer in a network definition.
Definition: NvInfer.h:6089
virtual ~IOneHotLayer() noexcept=default
apiv::VOneHotLayer * mImpl
Definition: NvInfer.h:6110
void setAxis(int32_t axis) noexcept
Set the axis parameter.
Definition: NvInfer.h:6096
int32_t getAxis() const noexcept
Get the value of the axis parameter.
Definition: NvInfer.h:6104
Optimization profile for dynamic input dimensions and shape tensors.
Definition: NvInferRuntime.h:2672
Layer that represents a padding operation.
Definition: NvInfer.h:2950
Dims getPostPaddingNd() const noexcept
Get the padding that is applied at the end of the tensor.
Definition: NvInfer.h:2999
void setPrePaddingNd(Dims const &padding) noexcept
Set the padding that is applied at the start of the tensor.
Definition: NvInfer.h:2961
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:2987
Dims getPrePaddingNd() const noexcept
Get the padding that is applied at the start of the tensor.
Definition: NvInfer.h:2973
apiv::VPaddingLayer * mImpl
Definition: NvInfer.h:3005
Layer that represents a parametric ReLU operation.
Definition: NvInfer.h:3968
apiv::VParametricReLULayer * mImpl
Definition: NvInfer.h:3970
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:139
Layer type for pluginV2.
Definition: NvInfer.h:2654
virtual ~IPluginV2Layer() noexcept=default
apiv::VPluginV2Layer * mImpl
Definition: NvInfer.h:2667
IPluginV2 & getPlugin() noexcept
Get the plugin for the layer.
Definition: NvInfer.h:2661
Layer type for V3 plugins.
Definition: NvInfer.h:2681
virtual ~IPluginV3Layer() noexcept=default
IPluginV3 & getPlugin() noexcept
Get the plugin for the layer.
Definition: NvInfer.h:2688
apiv::VPluginV3Layer * mImpl
Definition: NvInfer.h:2694
A Pooling layer in a network definition.
Definition: NvInfer.h:1416
PoolingType getPoolingType() const noexcept
Get the type of activation to be performed.
Definition: NvInfer.h:1435
PaddingMode getPaddingMode() const noexcept
Get the padding mode.
Definition: NvInfer.h:1568
Dims getPostPadding() const noexcept
Get the padding.
Definition: NvInfer.h:1544
bool getAverageCountExcludesPadding() const noexcept
Get whether average pooling uses as a denominator the overlap area between the window and the unpadde...
Definition: NvInfer.h:1488
Dims getPrePadding() const noexcept
Get the pre-padding.
Definition: NvInfer.h:1516
void setPoolingType(PoolingType type) noexcept
Set the type of activation to be performed.
Definition: NvInfer.h:1425
void setWindowSizeNd(Dims const &windowSize) noexcept
Set the multi-dimension window size for pooling.
Definition: NvInfer.h:1581
void setPaddingMode(PaddingMode paddingMode) noexcept
Set the padding mode.
Definition: NvInfer.h:1557
Dims getWindowSizeNd() const noexcept
Get the multi-dimension window size for pooling.
Definition: NvInfer.h:1591
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:1477
void setPaddingNd(Dims const &padding) noexcept
Set the multi-dimension padding for pooling.
Definition: NvInfer.h:1635
float getBlendFactor() const noexcept
Get the blending factor for the max_average_blend mode: max_average_blendPool = (1-blendFactor)*maxPo...
Definition: NvInfer.h:1463
void setStrideNd(Dims const &stride) noexcept
Set the multi-dimension stride for pooling.
Definition: NvInfer.h:1606
Dims getStrideNd() const noexcept
Get the multi-dimension stride for pooling.
Definition: NvInfer.h:1616
virtual ~IPoolingLayer() noexcept=default
Dims getPaddingNd() const noexcept
Get the multi-dimension padding for pooling.
Definition: NvInfer.h:1647
void setPostPadding(Dims const &padding) noexcept
Set the multi-dimension post-padding for pooling.
Definition: NvInfer.h:1534
void setPrePadding(Dims const &padding) noexcept
Set the multi-dimension pre-padding for pooling.
Definition: NvInfer.h:1506
void setBlendFactor(float blendFactor) noexcept
Set the blending factor for the max_average_blend mode: max_average_blendPool = (1-blendFactor)*maxPo...
Definition: NvInfer.h:1450
A Quantize layer in a network definition.
Definition: NvInfer.h:5465
void setToType(DataType toType) noexcept
Set the Quantize layer output type.
Definition: NvInfer.h:5526
bool setBlockShape(Dims const &blockShape) noexcept
Set the shape of the quantization block.
Definition: NvInfer.h:5499
void setAxis(int32_t axis) noexcept
Set the quantization axis.
Definition: NvInfer.h:5486
TRT_NODISCARD Dims getBlockShape() const noexcept
Get the shape of the quantization block.
Definition: NvInfer.h:5510
int32_t getAxis() const noexcept
Get the quantization axis.
Definition: NvInfer.h:5475
virtual ~IQuantizeLayer() noexcept=default
DataType getToType() const noexcept
Return the Quantize layer output type.
Definition: NvInfer.h:5538
A RaggedSoftmax layer in a network definition.
Definition: NvInfer.h:3802
apiv::VRaggedSoftMaxLayer * mImpl
Definition: NvInfer.h:3804
virtual ~IRaggedSoftMaxLayer() noexcept=default
A recurrence layer in a network definition.
Definition: NvInfer.h:4674
virtual ~IRecurrenceLayer() noexcept=default
Layer that represents a reduction across a non-bool tensor.
Definition: NvInfer.h:2870
void setKeepDimensions(bool keepDimensions) noexcept
Set the boolean that specifies whether or not to keep the reduced dimensions for the layer.
Definition: NvInfer.h:2917
void setOperation(ReduceOperation op) noexcept
Set the reduce operation for the layer.
Definition: NvInfer.h:2877
ReduceOperation getOperation() const noexcept
Get the reduce operation for the layer.
Definition: NvInfer.h:2887
virtual ~IReduceLayer() noexcept=default
uint32_t getReduceAxes() const noexcept
Get the axes over which to reduce for the layer.
Definition: NvInfer.h:2907
void setReduceAxes(uint32_t reduceAxes) noexcept
Set the axes over which to reduce.
Definition: NvInfer.h:2897
apiv::VReduceLayer * mImpl
Definition: NvInfer.h:2933
bool getKeepDimensions() const noexcept
Get the boolean that specifies whether or not to keep the reduced dimensions for the layer.
Definition: NvInfer.h:2927
A resize layer in a network definition.
Definition: NvInfer.h:4157
void setSelectorForSinglePixel(ResizeSelector selector) noexcept
Set coordinate selector function when resized to single pixel.
Definition: NvInfer.h:4318
void setNearestRounding(ResizeRoundMode value) noexcept
Set rounding mode for nearest neighbor resize.
Definition: NvInfer.h:4342
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:4236
void setOutputDimensions(Dims const &dimensions) noexcept
Set the output dimensions.
Definition: NvInfer.h:4177
void setCubicCoeff(float A) noexcept
Set the coefficient 'A' used in cubic interpolation.
Definition: NvInfer.h:4374
void setScales(float const *scales, int32_t nbScales) noexcept
Set the resize scales.
Definition: NvInfer.h:4217
float getCubicCoeff() const noexcept
Get the coefficient 'A' used in cubic interpolation.
Definition: NvInfer.h:4384
ResizeSelector getSelectorForSinglePixel() const noexcept
Get the coordinate selector function when resized to single pixel.
Definition: NvInfer.h:4328
InterpolationMode getResizeMode() const noexcept
Get resize mode for an input tensor.
Definition: NvInfer.h:4258
void setCoordinateTransformation(ResizeCoordinateTransformation coordTransform) noexcept
Set coordinate transformation function.
Definition: NvInfer.h:4293
void setExcludeOutside(bool excludeFlag) noexcept
Set the state for excluding outside pixels.
Definition: NvInfer.h:4397
void setResizeMode(InterpolationMode interpolationMode) noexcept
Set resize mode for an input tensor.
Definition: NvInfer.h:4248
Dims getOutputDimensions() const noexcept
Get the output dimensions.
Definition: NvInfer.h:4187
ResizeRoundMode getNearestRounding() const noexcept
Get rounding mode for nearest neighbor resize.
Definition: NvInfer.h:4352
bool getExcludeOutside() const noexcept
Get the state for excluding outside pixels.
Definition: NvInfer.h:4407
ResizeCoordinateTransformation getCoordinateTransformation() const noexcept
Get coordinate transformation function.
Definition: NvInfer.h:4303
A ReverseSequence layer in a network definition.
Definition: NvInfer.h:6395
void setSequenceAxis(int32_t sequenceAxis) noexcept
Set the sequence axis. Default is 0.
Definition: NvInfer.h:6428
int32_t getBatchAxis() const noexcept
Return the batch axis. Return 1 if no batch axis was set.
Definition: NvInfer.h:6415
apiv::VReverseSequenceLayer * mImpl
Definition: NvInfer.h:6444
int32_t getSequenceAxis() const noexcept
Return the sequence axis. Return 0 if no sequence axis was set.
Definition: NvInfer.h:6438
void setBatchAxis(int32_t batchAxis) noexcept
Set the batch axis. Default is 1.
Definition: NvInfer.h:6405
virtual ~IReverseSequenceLayer() noexcept=default
Layer that implements Rotary Position Embedding (RoPE) (https://arxiv.org/abs/2104....
Definition: NvInfer.h:7274
TRT_NODISCARD int32_t getRotaryEmbeddingDim() const noexcept
Get the number of hidden dimensions participating in RoPE. The default value is 0,...
Definition: NvInfer.h:7314
virtual ~IRotaryEmbeddingLayer() noexcept=default
void setInterleaved(bool interleaved) noexcept
Set whether the input is in interleaved format, i.e., whether the 2-d vectors rotated are taken from ...
Definition: NvInfer.h:7281
TRT_NODISCARD bool setRotaryEmbeddingDim(int32_t rotaryEmbeddingDim) noexcept
Set the number of hidden dimensions participating in RoPE. The default value is 0,...
Definition: NvInfer.h:7303
apiv::VRotaryEmbeddingLayer * mImpl
Definition: NvInfer.h:7337
TRT_NODISCARD bool getInterleaved() const noexcept
Get whether the input is in interleaved format. The default value is false.
Definition: NvInfer.h:7292
A Scale layer in a network definition.
Definition: NvInfer.h:1813
Weights getScale() const noexcept
Get the scale value.
Definition: NvInfer.h:1870
Weights getPower() const noexcept
Get the power value.
Definition: NvInfer.h:1890
void setScale(Weights scale) noexcept
Set the scale value.
Definition: NvInfer.h:1860
void setPower(Weights power) noexcept
Set the power value.
Definition: NvInfer.h:1880
ScaleMode getMode() const noexcept
Get the scale mode.
Definition: NvInfer.h:1830
void setShift(Weights shift) noexcept
Set the shift value.
Definition: NvInfer.h:1840
void setChannelAxis(int32_t channelAxis) noexcept
Set the channel axis.
Definition: NvInfer.h:1926
Weights getShift() const noexcept
Get the shift value.
Definition: NvInfer.h:1850
virtual ~IScaleLayer() noexcept=default
void setMode(ScaleMode mode) noexcept
Set the scale mode.
Definition: NvInfer.h:1820
int32_t getChannelAxis() const noexcept
Get the channel axis.
Definition: NvInfer.h:1905
A scatter layer in a network definition. Supports several kinds of scattering.
Definition: NvInfer.h:6017
void setMode(ScatterMode mode) noexcept
Set the scatter mode.
Definition: NvInfer.h:6024
apiv::VScatterLayer * mImpl
Definition: NvInfer.h:6058
void setAxis(int32_t axis) noexcept
Set the axis used by ScatterMode::kELEMENTS.
Definition: NvInfer.h:6044
int32_t getAxis() const noexcept
Get the axis.
Definition: NvInfer.h:6052
ScatterMode getMode() const noexcept
Get the scatter mode.
Definition: NvInfer.h:6034
virtual ~IScatterLayer() noexcept=default
Select elements from two data tensors based on a condition tensor.
Definition: NvInfer.h:4982
virtual ~ISelectLayer() noexcept=default
Layer type for getting shape of a tensor.
Definition: NvInfer.h:3475
virtual ~IShapeLayer() noexcept=default
apiv::VShapeLayer * mImpl
Definition: NvInfer.h:3477
Layer type for shuffling data.
Definition: NvInfer.h:3038
apiv::VShuffleLayer * mImpl
Definition: NvInfer.h:3196
void setFirstTranspose(Permutation permutation) noexcept
Set the permutation applied by the first transpose operation.
Definition: NvInfer.h:3049
void setSecondTranspose(Permutation permutation) noexcept
Set the permutation applied by the second transpose operation.
Definition: NvInfer.h:3149
Dims getReshapeDimensions() const noexcept
Get the reshaped dimensions.
Definition: NvInfer.h:3102
void setReshapeDimensions(Dims const &dimensions) noexcept
Set the reshaped dimensions.
Definition: NvInfer.h:3089
Permutation getFirstTranspose() const noexcept
Get the permutation applied by the first transpose operation.
Definition: NvInfer.h:3061
virtual ~IShuffleLayer() noexcept=default
Permutation getSecondTranspose() const noexcept
Get the permutation applied by the second transpose operation.
Definition: NvInfer.h:3161
bool getZeroIsPlaceholder() const noexcept
Get meaning of 0 in reshape dimensions.
Definition: NvInfer.h:3190
void setZeroIsPlaceholder(bool zeroIsPlaceholder) noexcept
Set meaning of 0 in reshape dimensions.
Definition: NvInfer.h:3177
Slices an input tensor into an output tensor based on the offset and strides.
Definition: NvInfer.h:3290
void setStride(Dims const &stride) noexcept
Set the stride for computing the output slice data.
Definition: NvInfer.h:3359
apiv::VSliceLayer * mImpl
Definition: NvInfer.h:3458
virtual ~ISliceLayer() noexcept=default
void setSize(Dims const &size) noexcept
Set the dimensions of the output slice.
Definition: NvInfer.h:3330
void setAxes(Dims const &axes) noexcept
Set the axes for this ISliceLayer.
Definition: NvInfer.h:3437
void setStart(Dims const &start) noexcept
Set the start offset that the slice layer uses to create the output slice.
Definition: NvInfer.h:3301
Dims getStart() const noexcept
Get the start offset for the slice layer.
Definition: NvInfer.h:3316
void setMode(SampleMode mode) noexcept
Set the slice mode.
Definition: NvInfer.h:3384
Dims getSize() const noexcept
Get dimensions of the output slice.
Definition: NvInfer.h:3345
SampleMode getMode() const noexcept
Get the slice mode.
Definition: NvInfer.h:3394
Dims getStride() const noexcept
Get the stride for the output slice.
Definition: NvInfer.h:3374
Dims getAxes() const noexcept
Get the axes for this ISliceLayer.
Definition: NvInfer.h:3452
A Softmax layer in a network definition.
Definition: NvInfer.h:1957
void setAxes(uint32_t axes) noexcept
Set the axis along which softmax is computed. Currently, only one axis can be set.
Definition: NvInfer.h:1979
uint32_t getAxes() const noexcept
Get the axis along which softmax occurs.
Definition: NvInfer.h:1989
virtual ~ISoftMaxLayer() noexcept=default
Layer that represents a squeeze operation, removing unit dimensions of the first input tensor on a se...
Definition: NvInfer.h:6607
virtual ~ISqueezeLayer() noexcept=default
apiv::VSqueezeLayer * mImpl
Definition: NvInfer.h:6624
A tensor in a network definition.
Definition: NvInfer.h:189
void setAllowedFormats(TensorFormats formats) noexcept
Set allowed formats for an input or output tensor. By default all formats are allowed....
Definition: NvInfer.h:340
void setDimensions(Dims const &dimensions) noexcept
Set the dimensions of a tensor.
Definition: NvInfer.h:237
void setName(char const *name) noexcept
Set the tensor name.
Definition: NvInfer.h:206
bool isExecutionTensor() const noexcept
Whether the tensor is an execution tensor.
Definition: NvInfer.h:405
char const * getName() const noexcept
Get the tensor name.
Definition: NvInfer.h:218
bool isShapeTensor() const noexcept
Whether the tensor is a shape tensor.
Definition: NvInfer.h:384
bool isNetworkInput() const noexcept
Whether the tensor is a network input.
Definition: NvInfer.h:310
TRT_DEPRECATED void setType(DataType type) noexcept
Set the data type of a tensor.
Definition: NvInfer.h:287
bool isNetworkOutput() const noexcept
Whether the tensor is a network output.
Definition: NvInfer.h:318
DataType getType() const noexcept
Get the data type of a tensor.
Definition: NvInfer.h:302
apiv::VTensor * mImpl
Definition: NvInfer.h:452
virtual ~ITensor() noexcept=default
void setDimensionName(int32_t index, char const *name) noexcept
Name a dimension of an input tensor.
Definition: NvInfer.h:431
char const * getDimensionName(int32_t index) const noexcept
Get the name of an input dimension.
Definition: NvInfer.h:446
Dims getDimensions() const noexcept
Get the dimensions of a tensor.
Definition: NvInfer.h:251
TensorFormats getAllowedFormats() const noexcept
Get a bitmask of TensorFormat values that the tensor supports. For a shape tensor,...
Definition: NvInfer.h:353
Class to handle tactic timing info collected from builder.
Definition: NvInfer.h:9874
int64_t queryKeys(TimingCacheKey *keyBuffer, int64_t capacity) const noexcept
Query cache keys from Timing Cache.
Definition: NvInfer.h:9940
bool combine(ITimingCache const &inputCache, bool ignoreMismatch) noexcept
Combine input timing cache into local instance.
Definition: NvInfer.h:9911
TimingCacheValue query(TimingCacheKey const &key) const noexcept
Query value in a cache entry.
Definition: NvInfer.h:9957
virtual ~ITimingCache() noexcept=default
bool update(TimingCacheKey const &key, TimingCacheValue const &value) noexcept
Update values in a cache entry.
Definition: NvInfer.h:9979
apiv::VTimingCache * mImpl
Definition: NvInfer.h:9985
bool reset() noexcept
Empty the timing cache.
Definition: NvInfer.h:9921
Layer that represents a TopK reduction.
Definition: NvInfer.h:3515
void setK(int32_t k) noexcept
Set the static k value for the layer.
Definition: NvInfer.h:3546
void setReduceAxes(uint32_t reduceAxes) noexcept
Set which axes to reduce for the layer.
Definition: NvInfer.h:3570
TopKOperation getOperation() const noexcept
Get the operation for the layer.
Definition: NvInfer.h:3532
apiv::VTopKLayer * mImpl
Definition: NvInfer.h:3629
void setOperation(TopKOperation op) noexcept
Set the operation for the layer.
Definition: NvInfer.h:3522
bool setIndicesType(DataType type) noexcept
Set the indices type for the layer.
Definition: NvInfer.h:3611
int32_t getK() const noexcept
Get the k value for the layer.
Definition: NvInfer.h:3560
uint32_t getReduceAxes() const noexcept
Get the axes to reduce for the layer.
Definition: NvInfer.h:3580
virtual ~ITopKLayer() noexcept=default
DataType getIndicesType() const noexcept
Return the TopK layer indices type.
Definition: NvInfer.h:3623
A layer that represents a trip-count limiter.
Definition: NvInfer.h:4795
TripLimit getTripLimit() const noexcept
Get a trip limiter type.
Definition: NvInfer.h:4800
virtual ~ITripLimitLayer() noexcept=default
Layer that represents an unary operation.
Definition: NvInfer.h:2762
void setOperation(UnaryOperation op) noexcept
Set the unary operation for the layer.
Definition: NvInfer.h:2771
apiv::VUnaryLayer * mImpl
Definition: NvInfer.h:2787
UnaryOperation getOperation() const noexcept
Get the unary operation for the layer.
Definition: NvInfer.h:2781
virtual ~IUnaryLayer() noexcept=default
Layer that represents an unsqueeze operation, which reshapes the first input tensor by inserting unit...
Definition: NvInfer.h:6637
virtual ~IUnsqueezeLayer() noexcept=default
apiv::VUnsqueezeLayer * mImpl
Definition: NvInfer.h:6655
An Interface class for version control.
Definition: NvInferRuntimeBase.h:279
Version information associated with a TRT interface.
Definition: NvInferRuntimeBase.h:244
An array of weights used as a layer parameter.
Definition: NvInferRuntime.h:124
Definition: NvInferRuntimeBase.h:416
Definition: NvInferRuntime.h:1656
Definition: NvInferPluginBase.h:206
Definition: NvInfer.h:10232
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
IProgressMonitor()=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.
Definition: NvInferRuntime.h:666
IBuilder * createInferBuilder(ILogger &logger) noexcept
Create an instance of an IBuilder class.
Definition: NvInfer.h:11500
The TensorRT API version 1 namespace.
Definition: NvInferPluginBase.h:29
uint32_t TacticSources
Represents a collection of one or more TacticSource values combine using bitwise-OR operations.
Definition: NvInferRuntime.h:2958
ResizeSelector
The coordinate selector when resize to single pixel output.
Definition: NvInfer.h:4062
@ kFORMULA
Use formula to map the original index.
@ kUPPER
Select the upper left pixel.
EngineCapability
List of supported engine capability flows.
Definition: NvInferRuntime.h:76
MemoryPoolType
The type for memory pools used by TensorRT.
Definition: NvInfer.h:9996
TENSORRTAPI bool setInternalLibraryPath(AsciiChar const *path) noexcept
Set a custom directory path for loading internal TensorRT libraries when building engines.
ScaleMode
Controls how shift, scale and power are applied in a Scale layer.
Definition: NvInfer.h:1770
@ kUNIFORM
Identical coefficients across all elements of the tensor.
@ kCHANNEL
Per-channel coefficients.
RuntimePlatform
Describes the intended runtime platform (operating system and CPU architecture) for the execution of ...
Definition: NvInfer.h:9569
@ kNONE
Tensor is not an input or output.
HardwareCompatibilityLevel
Describes requirements of compatibility with GPU architectures other than that of the GPU on which th...
Definition: NvInfer.h:10124
@ kSAME_COMPUTE_CAPABILITY
CumulativeOperation
Enumerates the cumulative operations that may be performed by a Cumulative layer.
Definition: NvInfer.h:6671
BoundingBoxFormat
Representation of bounding box data used for the Boxes input tensor in INMSLayer.
Definition: NvInfer.h:6209
@ 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:9808
constexpr int32_t EnumMax< LayerType >() noexcept
Definition: NvInfer.h:124
ComputeCapability
Describes compute capability that an engine will be built for.
Definition: NvInfer.h:10173
@ kSM120
Target NVIDIA Blackwell GPU architecture (SM 12.0).
@ kSM75
Target NVIDIA Turing GPU architecture (SM 7.5).
@ kSM80
Target NVIDIA Ampere GPU architecture (SM 8.0).
@ kCURRENT
Use the compute capability of the current GPU in the environment.
@ kSM89
Target NVIDIA Ada Lovelace GPU architecture (SM 8.9).
@ kSM86
Target NVIDIA Ampere GPU architecture (SM 8.6).
UnaryOperation
Enumerates the unary operations that may be performed by a Unary layer.
Definition: NvInfer.h:2715
@ 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.
@ 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.
constexpr int32_t EnumMax< ReduceOperation >() noexcept
Definition: NvInfer.h:2830
constexpr int32_t EnumMax< TripLimit >() noexcept
Definition: NvInfer.h:4463
ActivationType
Enumerates the types of activation to perform in an activation layer.
Definition: NvInfer.h:143
@ kSELU
Selu activation: x>0 ? beta * x : beta * (alpha*exp(x) - alpha)
@ 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:5043
ResizeRoundMode
The rounding mode for nearest neighbor resize.
Definition: NvInfer.h:4092
@ kHALF_DOWN
Round half down.
char_t AsciiChar
Definition: NvInferRuntimeBase.h:116
PaddingMode
Enumerates the modes of padding to perform in convolution, deconvolution and pooling layer,...
Definition: NvInfer.h:948
@ 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
Enum that describes kinds of trip limits.
Definition: NvInfer.h:4451
@ 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:11176
PreviewFeature
Define preview features.
Definition: NvInfer.h:10071
@ kRUNTIME_ACTIVATION_RESIZE_10_10
@ kMULTIDEVICE_RUNTIME_10_16
@ kALIASED_PLUGIN_IO_10_03
TilingOptimizationLevel
Define the optimization levels for Tiling.
Definition: NvInfer.h:10199
@ kFAST
Use a fast algorithm and heuristic based strategy. Slightly increases engine build time.
@ kFULL
Increase search space even wider. Significantly increases engine build time.
constexpr int32_t EnumMax< GatherMode >() noexcept
Definition: NvInfer.h:2474
DataType
The type of weights and tensors. The datatypes other than kBOOL, kINT32, and kINT64 are "activation d...
Definition: NvInferRuntimeBase.h:146
uint32_t BuilderFlags
Represents one or more BuilderFlag values using binary OR operations, e.g., 1U << BuilderFlag::kFP16 ...
Definition: NvInfer.h:9601
DeviceType
The device that this layer/network will execute on.
Definition: NvInferRuntime.h:1350
constexpr int32_t EnumMax< ScaleMode >() noexcept
Definition: NvInfer.h:1782
LayerType
The type values of layer classes.
Definition: NvInfer.h:58
@ kGRID_SAMPLE
Grid sample layer.
@ kRAGGED_SOFTMAX
Ragged softmax layer.
@ kDECONVOLUTION
Deconvolution layer.
@ kASSERTION
Assertion layer.
@ kATTENTION_INPUT
Attention Input.
@ kMATRIX_MULTIPLY
Matrix multiply layer.
@ kCONDITION
Condition layer.
@ kCUMULATIVE
Cumulative layer.
@ kCONDITIONAL_INPUT
Conditional Input layer.
@ kIDENTITY
Identity layer.
@ kNORMALIZATION
Normalization layer.
@ kQUANTIZE
Quantize layer.
@ kCONVOLUTION
Convolution layer.
@ kPARAMETRIC_RELU
Parametric ReLU layer.
@ kATTENTION_OUTPUT
Attention Output.
@ kUNSQUEEZE
Unsqueeze Layer.
@ kCONCATENATION
Concatenation layer.
@ kREVERSE_SEQUENCE
Reverse sequence layer.
@ kROTARY_EMBEDDING
Rotary Embedding layer.
@ kRECURRENCE
Loop Recurrence layer.
@ kDEQUANTIZE
Dequantize layer.
@ kPLUGIN_V3
PluginV3 layer.
@ kITERATOR
Loop Iterator layer.
@ kTRIP_LIMIT
Loop Trip limit layer.
@ kDYNAMIC_QUANTIZE
Dynamic Quantize layer.
@ kUNARY
UnaryOp operation Layer.
@ kACTIVATION
Activation layer.
@ kELEMENTWISE
Elementwise layer.
@ kPLUGIN_V2
PluginV2 layer.
@ kLOOP_OUTPUT
Loop output layer.
@ kCONDITIONAL_OUTPUT
Conditional Output layer.
@ kCONSTANT
Constant layer.
@ kNON_ZERO
NonZero layer.
@ kKVCACHE_UPDATE
KV Cache Update layer.
@ kDIST_COLLECTIVE
DistCollective layer.
SampleMode
Controls how ISliceLayer and IGridSample handle out-of-bounds coordinates.
Definition: NvInfer.h:3206
@ 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:2462
@ kDEFAULT
Similar to ONNX Gather.
@ kELEMENT
Similar to ONNX GatherElements.
@ kND
Similar to ONNX GatherND.
MoEActType
Enumerates the activation type for the MoE layer.
Definition: NvInfer.h:7435
uint32_t TensorFormats
It is capable of representing one or more TensorFormat by binary OR operations, e....
Definition: NvInfer.h:135
ProfilingVerbosity
List of verbosity levels of layer information exposed in NVTX annotations and in IEngineInspector.
Definition: NvInferRuntime.h:2970
NetworkDefinitionCreationFlag
List of immutable network properties expressed at network creation time. NetworkDefinitionCreationFla...
Definition: NvInfer.h:11187
ElementWiseOperation
Enumerates the binary operations that may be performed by an ElementWise layer.
Definition: NvInfer.h:2372
@ 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.
CollectiveOperation
Enumerates the collective operations that may be performed by a DistCollective layer.
Definition: NvInfer.h:2843
@ kREDUCE_SCATTER
Reduce scatter.
constexpr int32_t EnumMax< SampleMode >() noexcept
Definition: NvInfer.h:3222
InterpolationMode
Enumerates various modes of interpolation.
Definition: NvInfer.h:3980
@ 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:9611
@ 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.
@ kERROR_ON_TIMING_CACHE_MISS
@ kEDITABLE_TIMING_CACHE
Enable editable timing cache.
@ kPREFER_PRECISION_CONSTRAINTS
@ kDISTRIBUTIVE_INDEPENDENCE
@ kSTRIP_PLAN
Strip the refittable weights from the engine plan file.
@ kMONITOR_MEMORY
Enable memory monitor during build time.
@ kDISABLE_TIMING_CACHE
Disable reuse of timing information across identical layers.
@ kDISABLE_COMPILATION_CACHE
@ kREFIT
Enable building a refittable engine.
@ kOBEY_PRECISION_CONSTRAINTS
@ kREJECT_EMPTY_ALGORITHMS
constexpr int32_t EnumMax< TopKOperation >() noexcept
Definition: NvInfer.h:3498
TENSORRTAPI nvinfer1::IPluginRegistry * getBuilderPluginRegistry(nvinfer1::EngineCapability capability) noexcept
Return the plugin registry for building a Standard engine, or nullptr if no registry exists.
constexpr int32_t EnumMax< MemoryPoolType >() noexcept
Definition: NvInfer.h:10057
TopKOperation
Enumerates the operations that may be performed by a TopK layer.
Definition: NvInfer.h:3487
ReduceOperation
Enumerates the reduce operations that may be performed by a Reduce layer.
Definition: NvInfer.h:2815
@ kAVG
Average of the elements.
constexpr int32_t EnumMax< LoopOutput >() noexcept
Definition: NvInfer.h:4440
constexpr int32_t EnumMax< NetworkDefinitionCreationFlag >() noexcept
Definition: NvInfer.h:11206
ScatterMode
Control form of IScatterLayer.
Definition: NvInfer.h:5943
MatrixOperation
Enumerates the operations that may be performed on a tensor by IMatrixMultiplyLayer before multiplica...
Definition: NvInfer.h:3640
@ kTRANSPOSE
Like kNONE, but transpose the matrix dimensions.
ResizeCoordinateTransformation
The resize coordinate transformation function.
Definition: NvInfer.h:4008
constexpr int32_t EnumMax< UnaryOperation >() noexcept
Definition: NvInfer.h:2749
LoopOutput
Enum that describes kinds of loop outputs.
Definition: NvInfer.h:4423
@ 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:6222
constexpr int32_t EnumMax< MatrixOperation >() noexcept
Definition: NvInfer.h:3668
KVCacheMode
Enumerates the KVCache modes that may be performed by a KVCacheUpdate layer.
Definition: NvInfer.h:7347
PoolingType
The type of pooling to perform in a pooling layer.
Definition: NvInfer.h:1384
@ 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:10315
constexpr int32_t EnumMax< FillOperation >() noexcept
Definition: NvInfer.h:5077
AttentionNormalizationOp
Enumerates the operations that may be performed by the normalization in the attention subgraph.
Definition: NvInfer.h:6806
constexpr int32_t EnumMax< ScatterMode >() noexcept
Definition: NvInfer.h:5954
Represents a permutation of dimensions.
Definition: NvInfer.h:3015
Declaration of EnumMaxImpl struct to store maximum number of elements in an enumeration type.
Definition: NvInferRuntimeBase.h:129
The key to retrieve timing cache entries.
Definition: NvInfer.h:9834
Definition: NvInfer.h:9848
uint64_t tacticHash
Hash of the selected tactic.
Definition: NvInfer.h:9850
float timingMSec
Timing of this tactic in milliseconds. Negative numbers and NaN are invalid values.
Definition: NvInfer.h:9852