TensorRT-LLMs/cpp/tensorrt_llm/layers/topPSamplingLayer.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

312 lines
14 KiB
C++

/*
* Copyright (c) 2019-2024, NVIDIA CORPORATION. All rights reserved.
* Copyright (c) 2021, NAVER Corp. Authored by CLOVA.
*
* 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 "topPSamplingLayer.h"
#include "tensorrt_llm/common/logger.h"
#include "tensorrt_llm/common/memoryUtils.h"
#include "tensorrt_llm/common/nvtxUtils.h"
#include "tensorrt_llm/kernels/decodingCommon.h"
#include "tensorrt_llm/kernels/samplingTopKKernels.h"
#include "tensorrt_llm/kernels/samplingTopPKernels.h"
#include "tensorrt_llm/layers/defaultDecodingParams.h"
#include "tensorrt_llm/layers/layerUtils.h"
#include <algorithm>
#include <cfloat>
using namespace tensorrt_llm::common;
using namespace tensorrt_llm::kernels;
using namespace tensorrt_llm::runtime;
namespace tensorrt_llm::layers
{
template <typename T>
TopPSamplingLayer<T>::TopPSamplingLayer(DecoderDomain const& decoderDomain,
std::shared_ptr<BufferManager> bufferManager, bool isDeterministic, bool isAirTopP)
: BaseLayer(decoderDomain, bufferManager)
, mIsDeterministic(isDeterministic)
, mIsAirTopP(isAirTopP)
{
TLLM_LOG_TRACE("%s start", __PRETTY_FUNCTION__);
auto const deviceId = getDevice();
TLLM_CUDA_CHECK(cudaGetDeviceProperties(&mDeviceProp, deviceId));
allocateBuffer(mDecoderDomain.getBatchSize());
TLLM_LOG_TRACE("%s stop", __PRETTY_FUNCTION__);
}
template <typename T>
void TopPSamplingLayer<T>::allocateBuffer(SizeType32 batchSize)
{
TLLM_LOG_TRACE("%s start", __PRETTY_FUNCTION__);
if (!mIsAirTopP)
{
mWorkspaceSize = getTopPWorkspaceSize<T>(batchSize, mDecoderDomain.getVocabSizePadded());
}
else
{
mWorkspaceSize = getAirTopPWorkspaceSize<T>(batchSize, mDecoderDomain.getVocabSizePadded(), mIsDeterministic);
}
auto const batchSizeShape = ITensor::makeShape({batchSize});
mRuntimeTopKDevice = mBufferManager->gpu(batchSizeShape, TRTDataType<SizeType32>::value);
mRuntimeTopPDevice = mBufferManager->gpu(batchSizeShape, TRTDataType<float>::value);
mInitialTopPDevice = mBufferManager->gpu(batchSizeShape, TRTDataType<float>::value);
mTopPDecayDevice = mBufferManager->gpu(batchSizeShape, TRTDataType<float>::value);
mTopPMinDevice = mBufferManager->gpu(batchSizeShape, TRTDataType<float>::value);
mTopPResetIdsDevice = mBufferManager->gpu(batchSizeShape, TRTDataType<TokenIdType>::value);
mSkipDecodeDevice = mBufferManager->gpu(batchSizeShape, TRTDataType<bool>::value);
mSkipDecodeHost = mBufferManager->pinnedPool(batchSizeShape, TRTDataType<bool>::value);
auto skipDecodeHostRange = BufferRange<bool>(*mSkipDecodeHost);
std::fill(skipDecodeHostRange.begin(), skipDecodeHostRange.end(), true);
mSetupWorkspaceSize = std::max({mRuntimeTopKDevice->getSizeInBytes(), mRuntimeTopPDevice->getSizeInBytes(),
mInitialTopPDevice->getSizeInBytes(), mTopPDecayDevice->getSizeInBytes(), mTopPMinDevice->getSizeInBytes(),
mTopPResetIdsDevice->getSizeInBytes(), mSkipDecodeDevice->getSizeInBytes()});
TLLM_LOG_TRACE("%s stop", __PRETTY_FUNCTION__);
}
template <typename T>
void TopPSamplingLayer<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(TopPSamplingLayer_setup);
auto setupParams = std::dynamic_pointer_cast<SamplingSetupParams>(baseSetupParams);
auto constexpr defaultTopPDecay = DefaultDecodingParams::getTopPDecay();
auto constexpr defaultTopPMin = DefaultDecodingParams::getTopPMin(); // prevent TopP becoming 0.0
auto const* batchSlotsHostPtr = bufferCastOrNull<SizeType32>(batchSlots);
auto* skipDecodeHostPtr = bufferCastOrNull<bool>(mSkipDecodeHost);
if (!setupParams->runtimeTopP.has_value() || setupParams->runtimeTopP.value().empty())
{
// Fast path to disable TopP for slots
for (SizeType32 bi = 0; bi < batchSize; ++bi)
{
auto const bid = batchSlotsHostPtr[bi];
skipDecodeHostPtr[bid] = true;
}
auto const maxBatchSize = mDecoderDomain.getBatchSize();
auto skipDecodeHostSlice = IBuffer::slice(mSkipDecodeHost, 0, maxBatchSize);
mBufferManager->copy(*skipDecodeHostSlice, *mSkipDecodeDevice);
return;
}
auto runtimeTopK = setupParams->runtimeTopK.value_or(std::vector{DefaultDecodingParams::getTopK()});
auto runtimeTopP = setupParams->runtimeTopP.value();
auto decayVec = setupParams->topPDecay.value_or(std::vector{defaultTopPDecay});
auto topPMinVec = setupParams->topPMin.value_or(std::vector{defaultTopPMin});
auto topPResetIdsVec = setupParams->topPResetIds.value_or(std::vector{DefaultDecodingParams::getTopPResetId()});
auto const paramsSize
= expandMatchElements(batchSize, runtimeTopK, runtimeTopP, decayVec, topPMinVec, topPResetIdsVec);
TLLM_CHECK_WITH_INFO(paramsSize != 0,
fmtstr("TopPSamplingLayer got parameter with unexpected size, want 1 or batchSize(%d), got"
"runtimeTopK.size() = %zu, "
"runtimeTopP.size() = %zu, "
"topPDecay.size() = %zu, "
"topPMin.size() = %zu, "
"topPResetIds.size() = %zu",
batchSize, runtimeTopK.size(), runtimeTopP.size(), decayVec.size(), topPMinVec.size(),
topPResetIdsVec.size()));
for (size_t i = 0; i < paramsSize; ++i)
{
// support topK up to TOP_K_MAX.
auto& topK = runtimeTopK[i];
auto& topP = runtimeTopP[i];
clampTopK(topK);
clampTopP(topP);
regularizeTopKTopP(topK, topP);
auto& decay = decayVec[i];
if (decay <= 0.f || decay > 1.0f)
{
TLLM_LOG_WARNING(
"Decay (%f) is out of range ((0.0, 1.0f]). Change to default (%f).", decay, defaultTopPDecay);
decay = defaultTopPDecay;
}
auto& topPMin = topPMinVec[i];
if (topPMin <= 0.f || topPMin > 1.0f)
{
TLLM_LOG_WARNING(
"TopP min (%f) is out of range ([0.0, 1.0f]). Change to default (%f).", topPMin, defaultTopPMin);
topPMin = defaultTopPMin;
}
}
// Update parameters on both device and host, so we can
// determine whether we can skip launch kernel by examine mSkipDecodeHost
// without consulting device memory, or we'll have to do an expensive synchronization.
SizeType32* topKsPtr = nullptr;
float* topPsPtr = nullptr;
float* topPDecayPtr = nullptr;
float* topPMinPtr = nullptr;
SizeType32* topPResetIdsPtr = nullptr;
if (paramsSize > 1)
{
auto initWorkspaceSizes = getTopPInitWorkspaceSizes(batchSize);
std::vector<void*> alignedPointers;
calcAlignedPointers(workspace->getRawWorkspaceDevicePtr(), initWorkspaceSizes)(
topKsPtr, topPsPtr, topPDecayPtr, topPMinPtr, topPResetIdsPtr);
DecodingLayerWorkspace::copyToWorkspace(
*mBufferManager, runtimeTopK, IBuffer::wrap(topKsPtr, initWorkspaceSizes[0] / sizeof(*topKsPtr)));
DecodingLayerWorkspace::copyToWorkspace(
*mBufferManager, runtimeTopP, IBuffer::wrap(topPsPtr, initWorkspaceSizes[1] / sizeof(*topPsPtr)));
DecodingLayerWorkspace::copyToWorkspace(
*mBufferManager, decayVec, IBuffer::wrap(topPDecayPtr, initWorkspaceSizes[2] / sizeof(*topPDecayPtr)));
DecodingLayerWorkspace::copyToWorkspace(
*mBufferManager, topPMinVec, IBuffer::wrap(topPMinPtr, initWorkspaceSizes[3] / sizeof(*topPMinPtr)));
DecodingLayerWorkspace::copyToWorkspace(*mBufferManager, topPResetIdsVec,
IBuffer::wrap(topPResetIdsPtr, initWorkspaceSizes[4] / sizeof(*topPResetIdsPtr)));
}
auto const* batchSlotsDevicePtr = workspace->getDeviceBatchSlotsPtr();
auto* skipDecodeDevicePtr = bufferCastOrNull<bool>(mSkipDecodeDevice);
auto* initialTopPDevicePtr = bufferCast<float>(*mInitialTopPDevice);
invokeSetTopPRuntimeArgs(batchSize, //
{topKsPtr, runtimeTopK.front(), bufferCast<SizeType32>(*mRuntimeTopKDevice)}, //
{topPsPtr, runtimeTopP.front(), bufferCast<float>(*mRuntimeTopPDevice)}, //
skipDecodeDevicePtr, initialTopPDevicePtr, batchSlotsDevicePtr, true, getStream());
invokeScatterDecodingParams(topPDecayPtr, decayVec.front(), bufferCast<float>(*mTopPDecayDevice),
batchSlotsDevicePtr, batchSize, getStream());
invokeScatterDecodingParams(topPMinPtr, topPMinVec.front(), bufferCast<float>(*mTopPMinDevice), batchSlotsDevicePtr,
batchSize, getStream());
invokeScatterDecodingParams(topPResetIdsPtr, topPResetIdsVec.front(), bufferCast<TokenIdType>(*mTopPResetIdsDevice),
batchSlotsDevicePtr, batchSize, getStream());
topKsPtr = paramsSize > 1 ? runtimeTopK.data() : nullptr;
invokeSetTopPRuntimeArgs(batchSize, //
{topKsPtr, runtimeTopK.front(), nullptr}, {}, //
skipDecodeHostPtr, nullptr, batchSlotsHostPtr, false);
if (mIsAirTopP)
{
auto smCnt = mDeviceProp.multiProcessorCount;
if (smCnt <= 0)
{
auto const deviceId = getDevice();
cudaDeviceProp prop{};
TLLM_CUDA_CHECK(cudaGetDeviceProperties(&prop, deviceId));
smCnt = prop.multiProcessorCount;
}
mAirTopPBlockNum
= calcAirTopPBlockNum<T>(batchSize, mDecoderDomain.getVocabSizePadded(), smCnt, mIsDeterministic);
}
TLLM_LOG_TRACE("%s stop", __PRETTY_FUNCTION__);
}
template <typename T>
void TopPSamplingLayer<T>::forwardAsync(std::shared_ptr<BaseDecodingOutputs> const& outputs,
std::shared_ptr<BaseDecodingInputs> const& baseInputs,
std::shared_ptr<runtime::DecodingLayerWorkspace> const& workspace)
{
TLLM_LOG_TRACE("%s start", __PRETTY_FUNCTION__);
NVTX3_SCOPED_RANGE(TopPSamplingLayer_forwardAsync);
auto inputs = std::dynamic_pointer_cast<SamplingInputs>(baseInputs);
auto const batchSize = inputs->logits.value()->getDimension<0>();
auto const* batchSlotsHost = bufferCast<SizeType32>(*inputs->batchSlots);
auto* skipDecodeHostPtr = bufferCastOrNull<bool>(mSkipDecodeHost);
auto const skip = allOfBatchSlots(batchSlotsHost, skipDecodeHostPtr, batchSize, true);
if (skip)
{
return;
}
// Probabilities must be already computed instead of logits
auto probs = bufferCastOrNull<T>(inputs->logits);
auto const* endIds = bufferCastOrNull<TokenIdType>(inputs->endIds);
auto const* finishedInput = (inputs->finished) ? reinterpret_cast<FinishedState const*>(
bufferCastOrNull<FinishedState::UnderlyingType>(inputs->finished.value()))
: nullptr;
auto* finishedOutput = (outputs->finished)
? reinterpret_cast<FinishedState*>(bufferCastOrNull<FinishedState::UnderlyingType>(outputs->finished.value()))
: nullptr;
auto* cumLogProbs = bufferCastOrNull<float>(outputs->cumLogProbs);
auto* outputLogProbs = bufferCastOrNull<float>(outputs->outputLogProbsTiled);
auto* sequenceLength = bufferCastOrNull<SizeType32>(outputs->sequenceLength);
TopPSamplingKernelParams<T> params{};
params.probs = probs;
params.outputIdsPtrs = bufferCastOrNull<TokenIdType*>(outputs->outputIdsPtr);
params.workspace = workspace->getRawWorkspaceDevicePtr();
params.topPs = bufferCastOrNull<float>(mRuntimeTopPDevice);
params.sequenceLength = sequenceLength;
params.endIds = endIds;
params.batchSlots = workspace->getDeviceBatchSlotsPtr();
params.finishedInput = finishedInput;
params.finishedOutput = finishedOutput;
params.skipDecode = bufferCastOrNull<bool>(mSkipDecodeDevice);
params.cumLogProbs = cumLogProbs;
params.outputLogProbs = outputLogProbs;
params.curandState = inputs->curandStates;
params.batchSize = batchSize;
params.maxBatchSize = mDecoderDomain.getBatchSize();
params.vocabSizePadded = mDecoderDomain.getVocabSizePadded();
if (!mIsAirTopP)
{
invokeBatchTopPSampling<T>(params, getStream());
}
else
{
params.blockNum = mAirTopPBlockNum;
params.isDeterministic = mIsDeterministic;
invokeBatchAirTopPSampling<T>(params, getStream());
}
sync_check_cuda_error(getStream());
auto* runtimeTopPDevicePtr = bufferCastOrNull<float>(mRuntimeTopPDevice);
auto* initialTopPDevicePtr = bufferCastOrNull<float>(mInitialTopPDevice);
auto* topPDecayDevicePtr = bufferCastOrNull<float>(mTopPDecayDevice);
auto* topPMinDevicePtr = bufferCastOrNull<float>(mTopPMinDevice);
auto* topPResetIdsDevice = bufferCastOrNull<TokenIdType>(mTopPResetIdsDevice);
auto* outputIdsPtr = bufferCastOrNull<TokenIdType const*>(outputs->outputIdsPtr);
invokeComputeToppDecay(runtimeTopPDevicePtr, initialTopPDevicePtr, outputIdsPtr, topPDecayDevicePtr,
topPMinDevicePtr, topPResetIdsDevice, sequenceLength, workspace->getDeviceBatchSlotsPtr(), batchSize,
getStream());
TLLM_LOG_TRACE("%s stop", __PRETTY_FUNCTION__);
}
template <typename T>
size_t TopPSamplingLayer<T>::getWorkspaceSize() const noexcept
{
return std::max(mSetupWorkspaceSize, mWorkspaceSize);
}
template class TopPSamplingLayer<float>;
template class TopPSamplingLayer<half>;
} // namespace tensorrt_llm::layers