TensorRT-LLMs/cpp/tensorrt_llm/layers/explicitDraftTokensLayer.cpp
Robin Kobus 6d4b045d1f
refactor: Remove enforced sorted order of batch slots (#3502)
Signed-off-by: Robin Kobus <19427718+Funatiq@users.noreply.github.com>
2025-07-14 17:23:02 +02:00

303 lines
14 KiB
C++

/*
* Copyright (c) 2024, NVIDIA CORPORATION. All rights reserved.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include "explicitDraftTokensLayer.h"
#include "tensorrt_llm/common/nvtxUtils.h"
#include "tensorrt_llm/kernels/penaltyTypes.h"
#include "tensorrt_llm/kernels/speculativeDecoding/common.h"
#include "tensorrt_llm/kernels/speculativeDecoding/explicitDraftTokensKernels.h"
#include "tensorrt_llm/layers/defaultDecodingParams.h"
#include "tensorrt_llm/layers/layerUtils.h"
using namespace tensorrt_llm::common;
using namespace tensorrt_llm::kernels;
using namespace tensorrt_llm::kernels::speculative_decoding;
using namespace tensorrt_llm::runtime;
namespace tensorrt_llm::layers
{
template <typename T>
ExplicitDraftTokensLayer<T>::ExplicitDraftTokensLayer(
DecoderDomain const& decoderDomain, std::shared_ptr<BufferManager> bufferManager)
: BaseLayer(decoderDomain, bufferManager)
{
TLLM_LOG_TRACE("%s start", __PRETTY_FUNCTION__);
allocateBuffer();
TLLM_LOG_TRACE("%s stop", __PRETTY_FUNCTION__);
}
template <typename T>
void ExplicitDraftTokensLayer<T>::allocateBuffer()
{
TLLM_LOG_TRACE("%s start", __PRETTY_FUNCTION__);
mTemperature
= mBufferManager->pinnedPool(ITensor::makeShape({mDecoderDomain.getBatchSize()}), TRTDataType<float>::value);
mWorkspaceSize = invokeScanReduceGenerationLengths(
mDecoderDomain.getBatchSize(), nullptr, nullptr, 0, nullptr, nullptr, getStream());
mCurandStatesDevice = mBufferManager->gpu(
ITensor::makeShape({mDecoderDomain.getBatchSize(), sizeof(curandState_t)}), TRTDataType<int8_t>::value);
auto const batchSizeShape = ITensor::makeShape({mDecoderDomain.getBatchSize()});
mGenerationLengthInclusiveSum = mBufferManager->gpu(batchSizeShape, TRTDataType<SizeType32>::value);
mMaxGenerationLength = mBufferManager->gpu(ITensor::makeShape({1}), TRTDataType<SizeType32>::value);
mTemperatureDevice = mBufferManager->gpu(batchSizeShape, TRTDataType<float>::value);
mBestPathIndicesSlots = mBufferManager->gpu(batchSizeShape, TRTDataType<SizeType32>::value);
mLastDraftIndicesSlots = mBufferManager->gpu(ITensor::makeShape({mDecoderDomain.getBatchSize()
* mDecoderDomain.getSpeculativeDecodingModule()->getMaxNumPaths()
* mDecoderDomain.getSpeculativeDecodingModule()->getMaxPathLen()}),
TRTDataType<SizeType32>::value);
TLLM_LOG_TRACE("%s stop", __PRETTY_FUNCTION__);
}
template <typename T>
void ExplicitDraftTokensLayer<T>::setup(SizeType32 batchSize, SizeType32 beamWidth, TensorConstPtr batchSlots,
std::shared_ptr<BaseSetupParams> const& baseSetupParams,
std::shared_ptr<runtime::DecodingLayerWorkspace> const& workspace)
{
TLLM_LOG_TRACE("%s start", __PRETTY_FUNCTION__);
NVTX3_SCOPED_RANGE(ExplicitDraftTokensLayer_setup);
auto setupParams = std::dynamic_pointer_cast<ExplicitDraftTokensSetupParams>(baseSetupParams);
workspace->initializeDeviceCurandStates(
setupParams->randomSeed, batchSize, workspace->getDeviceBatchSlots(), mCurandStatesDevice);
// Setup penalties.
FillBuffers const fillBuffers{batchSize, mDecoderDomain.getBatchSize(), mBufferManager};
// Set decoder dtype to WAR the lack of bf16 support in decoder.
if (!mDecoderDtype)
{
mDecoderDtype = setupParams->dtype;
}
fillBuffers(setupParams->temperature, DefaultDecodingParams::getTemperature(), mTemperature, mTemperatureDevice,
batchSlots, getLimitsPenalty(DecodingPenaltyType::Temperature), "temperature penalty");
// Dispatch context buffer fill
if (mDecoderDtype == nvinfer1::DataType::kFLOAT)
{
fillContextBuffers<float>(batchSize, batchSlots, *setupParams, workspace);
}
else if (mDecoderDtype == nvinfer1::DataType::kHALF)
{
fillContextBuffers<half>(batchSize, batchSlots, *setupParams, workspace);
}
else if (mDecoderDtype == nvinfer1::DataType::kBF16)
{
fillContextBuffers<__nv_bfloat16>(batchSize, batchSlots, *setupParams, workspace);
}
TLLM_LOG_TRACE("%s stop", __PRETTY_FUNCTION__);
}
template <typename T>
void ExplicitDraftTokensLayer<T>::forwardAsync(std::shared_ptr<BaseDecodingOutputs> const& baseOutputs,
std::shared_ptr<BaseDecodingInputs> const& baseInputs,
std::shared_ptr<runtime::DecodingLayerWorkspace> const& workspace)
{
TLLM_LOG_TRACE("%s start", __PRETTY_FUNCTION__);
NVTX3_SCOPED_RANGE(ExplicitDraftTokensLayer_forwardAsync);
auto inputs = std::dynamic_pointer_cast<ExplicitDraftTokensInputs>(baseInputs);
auto outputs = std::dynamic_pointer_cast<ExplicitDraftTokensOutputs>(baseOutputs);
// DO NOT CHANGE THE ORDER.
// Convert masks to packed masks per request.
convertPackedMask(*outputs, *inputs, workspace);
// Slice output ids, pos ids, next draft tokens.
if (mDecoderDtype == nvinfer1::DataType::kFLOAT)
{
splitInputDataToBatchSlots<float>(*outputs, *inputs, workspace);
}
else if (mDecoderDtype == nvinfer1::DataType::kHALF)
{
splitInputDataToBatchSlots<half>(*outputs, *inputs, workspace);
}
else if (mDecoderDtype == nvinfer1::DataType::kBF16)
{
splitInputDataToBatchSlots<__nv_bfloat16>(*outputs, *inputs, workspace);
}
// Pack accepted paths for KV cache rewind.
packAcceptedPaths(*outputs, *inputs, workspace);
TLLM_LOG_TRACE("%s stop", __PRETTY_FUNCTION__);
}
template <typename T>
size_t ExplicitDraftTokensLayer<T>::getWorkspaceSize() const noexcept
{
return mWorkspaceSize;
}
template <typename T>
template <typename Dtype>
void ExplicitDraftTokensLayer<T>::fillContextBuffers(SizeType32 batchSize, BufferConstPtr batchSlots,
ExplicitDraftTokensSetupParams const& setupParams,
std::shared_ptr<runtime::DecodingLayerWorkspace> const& workspace)
{
TLLM_LOG_TRACE("%s start", __PRETTY_FUNCTION__);
FillContextExplicitDraftTokensParams<Dtype> params;
params.randDataSample = bufferCast<Dtype>(*setupParams.randomDataSample);
params.outputTemperatures = bufferCast<Dtype>(*setupParams.temperatures);
params.inputTemperatures = bufferCastOrNull<float>(mTemperatureDevice);
params.curandState = reinterpret_cast<curandState_t*>(bufferCastOrNull<int8_t>(mCurandStatesDevice));
params.batchSlots = workspace->getDeviceBatchSlotsPtr();
params.batchSize = batchSize;
params.checkParams();
invokeFillContextBuffers(params, getStream());
TLLM_LOG_TRACE("%s stop", __PRETTY_FUNCTION__);
}
template <typename T>
template <typename Dtype>
void ExplicitDraftTokensLayer<T>::splitInputDataToBatchSlots(ExplicitDraftTokensOutputs const& outputs,
ExplicitDraftTokensInputs const& inputs, std::shared_ptr<runtime::DecodingLayerWorkspace> const& workspace)
{
TLLM_LOG_TRACE("%s start", __PRETTY_FUNCTION__);
auto const batchSize = inputs.localBatchSize;
auto const maxSeqLen = outputs.outputIds->getDimension<-1>();
ExtractExplicitDraftTokensParams<Dtype> params;
params.outputIds = bufferCast<TokenIdType>(*outputs.outputIds);
params.outputPositionIdsBase = bufferCast<SizeType32>(*outputs.positionIdsBase);
params.outputPositionIds = bufferCast<SizeType32>(*outputs.nextDraftPosIds);
params.outputNextDraftTokens = bufferCast<TokenIdType>(*outputs.nextDraftTokens);
params.unpackedNextDraftTokens = bufferCast<TokenIdType>(*outputs.unpackedNextDraftTokens);
params.unpackedNextDraftIndices = bufferCast<SizeType32>(*outputs.unpackedNextDraftIndices);
params.acceptedLengths = bufferCast<SizeType32>(*outputs.numNewTokens.value());
params.nextDraftLengths = bufferCast<SizeType32>(*outputs.nextDraftLengths);
params.prevDraftLengths = bufferCast<SizeType32>(*outputs.prevDraftLengths);
params.sequenceLengths = bufferCast<SizeType32>(*outputs.sequenceLength.value());
params.randDataSample = bufferCast<Dtype>(*outputs.randomDataSample);
params.randDataVerification = bufferCast<Dtype>(*outputs.randomDataValidation);
params.outputDraftProbs = bufferCast<Dtype>(*outputs.nextDraftProbs);
params.outputTemperatures = bufferCast<Dtype>(*outputs.temperatures);
params.outputGenerationLengths = bufferCast<SizeType32>(*outputs.generationLengths);
params.outputBestPathIndices = bufferCast<SizeType32>(*mBestPathIndicesSlots);
params.outputLastDraftIndices = bufferCast<SizeType32>(*mLastDraftIndicesSlots);
params.batchSlots = bufferCast<SizeType32>(*inputs.seqSlots);
params.nextDraftTokens = bufferCast<TokenIdType>(*inputs.nextDraftTokens);
params.lastDraftTokens = bufferCast<TokenIdType>(*inputs.lastDraftTokens);
params.inputUnpackedNextDraftIndices = bufferCast<SizeType32>(*inputs.nextDraftIndices);
params.bestPathLengths = bufferCast<SizeType32>(*inputs.bestPathLengths);
params.bestPathIndices = bufferCast<SizeType32>(*inputs.bestPathIndices);
params.inputPositionIdsBase = bufferCast<SizeType32>(*inputs.positionIdsBase);
params.packedPositionIds = bufferCast<SizeType32>(*inputs.packedPosIds);
params.nextFlatTokens = bufferCast<TokenIdType>(*inputs.nextFlatTokens);
params.nextDraftProbs = bufferCast<Dtype>(*inputs.nextDraftProbs);
params.lastGenerationLengths = bufferCastOrNull<SizeType32>(inputs.lastGenerationLengths);
params.generationLengthInclusiveSum = bufferCast<SizeType32>(*mGenerationLengthInclusiveSum);
params.lastDraftIndices = bufferCast<SizeType32>(*inputs.lastDraftIndices);
params.inputTemperatures = bufferCast<float>(*mTemperatureDevice);
params.curandState = reinterpret_cast<curandState_t*>(bufferCastOrNull<int8_t>(mCurandStatesDevice));
params.batchSize = batchSize;
params.numPaths = mDecoderDomain.getSpeculativeDecodingModule()->getMaxNumPaths();
params.maxPathLength = mDecoderDomain.getSpeculativeDecodingModule()->getMaxPathLen();
params.maxSeqLen = maxSeqLen;
params.vocabSize = mDecoderDomain.getVocabSizePadded();
params.numContextRequests = batchSize - inputs.lastDraftTokens->getDimension<0>();
params.numGenerationRequests = inputs.lastDraftTokens->getDimension<0>();
params.checkParams();
// Copy max generation length
mBufferManager->copy(*inputs.maxGenLengthDevice, *outputs.maxGenLengthHost);
invokeExtractExplicitDraftTokens(params, getStream());
invokeCopyProbs(params, getStream());
// Copy generation lengths
mBufferManager->copy(*outputs.generationLengths, *outputs.generationLengthsHost);
TLLM_LOG_TRACE("%s stop", __PRETTY_FUNCTION__);
}
template <typename T>
void ExplicitDraftTokensLayer<T>::convertPackedMask(ExplicitDraftTokensOutputs const& outputs,
ExplicitDraftTokensInputs const& inputs, std::shared_ptr<runtime::DecodingLayerWorkspace> const& workspace)
{
TLLM_LOG_TRACE("%s start", __PRETTY_FUNCTION__);
auto batchSlots = bufferCast<SizeType32>(*inputs.seqSlots);
auto masksDevice = bufferCast<bool>(*inputs.masks);
auto generationLengths = bufferCast<SizeType32>(*inputs.generationLengths);
auto packedMasksDevice = bufferCast<SizeType32>(*outputs.packedMasks);
auto const batchSize = inputs.localBatchSize;
auto generationLengthInclusiveSumPtr = bufferCastOrNull<SizeType32>(mGenerationLengthInclusiveSum);
auto workSpaceDevicePtr = workspace->getRawWorkspaceDevicePtr();
auto maxGenerationLengthPtr = bufferCastOrNull<SizeType32>(mMaxGenerationLength);
invokeScanReduceGenerationLengths(batchSize, generationLengths, workSpaceDevicePtr, mWorkspaceSize,
generationLengthInclusiveSumPtr, maxGenerationLengthPtr, getStream());
invokeConvertMaskToPackedMask(batchSize, generationLengthInclusiveSumPtr, maxGenerationLengthPtr, masksDevice,
batchSlots, mDecoderDomain.getSpeculativeDecodingModule()->getMaxDecodingDraftTokens(),
mDecoderDomain.getSpeculativeDecodingModule()->getMaxDecodingTokens(), packedMasksDevice, getStream());
TLLM_LOG_TRACE("%s stop", __PRETTY_FUNCTION__);
}
template <typename T>
void ExplicitDraftTokensLayer<T>::packAcceptedPaths(ExplicitDraftTokensOutputs const& outputs,
ExplicitDraftTokensInputs const& inputs, std::shared_ptr<runtime::DecodingLayerWorkspace> const& workspace)
{
TLLM_LOG_TRACE("%s start", __PRETTY_FUNCTION__);
auto const batchSize = inputs.localBatchSize;
auto numNewTokens = bufferCast<SizeType32>(*outputs.numNewTokens.value());
auto numNewTokensCumSum = bufferCast<SizeType32>(*outputs.numNewTokensCumSum);
auto pathsOffsets = bufferCast<SizeType32>(*outputs.pathsOffsets);
auto batchSlots = workspace->getDeviceBatchSlotsPtr();
auto bestPathIndicesSlotsPtr = bufferCastOrNull<SizeType32>(mBestPathIndicesSlots);
auto lastDraftIndicesSlotsPtr = bufferCastOrNull<SizeType32>(mLastDraftIndicesSlots);
TLLM_CHECK_WITH_INFO(batchSlots != nullptr, "Batch slots must be provided for ExplicitDraftTokensLayer");
TLLM_CHECK_WITH_INFO(numNewTokens != nullptr, "Accepted lengths must be provided for ExplicitDraftTokensLayer");
TLLM_CHECK_WITH_INFO(
numNewTokensCumSum != nullptr, "numNewTokensCumSum must be provided for ExplicitDraftTokensLayer");
TLLM_CHECK_WITH_INFO(pathsOffsets != nullptr, "pathsOffsets must be provided for ExplicitDraftTokensLayer");
invokePackAcceptedPaths(numNewTokensCumSum, pathsOffsets, numNewTokens, bestPathIndicesSlotsPtr,
lastDraftIndicesSlotsPtr, batchSlots, batchSize, batchSize,
mDecoderDomain.getSpeculativeDecodingModule()->getMaxNumPaths(),
mDecoderDomain.getSpeculativeDecodingModule()->getMaxPathLen(), false, getStream());
TLLM_LOG_TRACE("%s stop", __PRETTY_FUNCTION__);
}
template class ExplicitDraftTokensLayer<float>;
template class ExplicitDraftTokensLayer<half>;
} // namespace tensorrt_llm::layers